Springer Lecture Notes in Physics 716
Springer Lecture Notes in Physics 716
Springer Lecture Notes in Physics 716
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
7<br />
Field-Theory Approaches<br />
to Nonequilibrium Dynamics<br />
U. C. Täuber<br />
Department of <strong>Physics</strong>, Center for Stochastic Processes <strong>in</strong> Science and Eng<strong>in</strong>eer<strong>in</strong>g<br />
Virg<strong>in</strong>ia Polytechnic Institute and State University<br />
Blacksburg, Virg<strong>in</strong>ia 24061-0435, USA<br />
tauber@vt.edu<br />
It is expla<strong>in</strong>ed how field-theoretic methods and the dynamic renormalisation<br />
group (RG) can be applied to study the universal scal<strong>in</strong>g properties of systems<br />
that either undergo a cont<strong>in</strong>uous phase transition or display generic scale <strong>in</strong>variance,<br />
both near and far from thermal equilibrium. Part 1 <strong>in</strong>troduces the<br />
response functional field theory representation of (nonl<strong>in</strong>ear) Langev<strong>in</strong> equations.<br />
The RG is employed to compute the scal<strong>in</strong>g exponents for several universality<br />
classes govern<strong>in</strong>g the critical dynamics near second-order phase transitions<br />
<strong>in</strong> equilibrium. The effects of reversible mode-coupl<strong>in</strong>g terms, quench<strong>in</strong>g<br />
from random <strong>in</strong>itial conditions to the critical po<strong>in</strong>t, and violat<strong>in</strong>g the<br />
detailed balance constra<strong>in</strong>ts are briefly discussed. It is shown how the same<br />
formalism can be applied to nonequilibrium systems such as driven diffusive<br />
lattice gases. Part 2 describes how the master equation for stochastic particle<br />
reaction processes can be mapped onto a field theory action. The RG is<br />
then used to analyse simple diffusion-limited annihilation reactions as well as<br />
generic cont<strong>in</strong>uous transitions from active to <strong>in</strong>active, absorb<strong>in</strong>g states, which<br />
are characterised by the power laws of (critical) directed percolation. Certa<strong>in</strong><br />
other important universality classes are mentioned, and some open issues are<br />
listed.<br />
7.1 Critical Dynamics<br />
Field-theoretic tools and the renormalisation group (RG) method have had<br />
a tremendous impact <strong>in</strong> our understand<strong>in</strong>g of the universal power laws that<br />
emerge near equilibrium critical po<strong>in</strong>ts (see, e.g., Refs. [1–6]), <strong>in</strong>clud<strong>in</strong>g the<br />
associated dynamic critical phenomena [7,8]. Our goal here is to similarly describe<br />
the scal<strong>in</strong>g properties of systems driven far from thermal equilibrium,<br />
which either undergo a cont<strong>in</strong>uous nonequilibrium phase transition or display<br />
generic scale <strong>in</strong>variance. We are then confronted with captur<strong>in</strong>g the (stochastic)<br />
dynamics of the long-wavelength modes of the “slow” degrees of freedom,<br />
U.C. Täuber: Field-Theory Approaches to Nonequilibrium Dynamics, Lect. <strong>Notes</strong> Phys. <strong>716</strong>,<br />
295–348 (2007)<br />
DOI 10.1007/3-540-69684-9 7 c○ <strong>Spr<strong>in</strong>ger</strong>-Verlag Berl<strong>in</strong> Heidelberg 2007
296 U. C. Täuber<br />
namely the order parameter for the transition, any conserved quantities, and<br />
perhaps additional relevant variables. In these lecture notes, I aim to briefly<br />
describe how a representation <strong>in</strong> terms of a field theory action can be obta<strong>in</strong>ed<br />
for (1) general nonl<strong>in</strong>ear Langev<strong>in</strong> stochastic differential equations [8, 9]; and<br />
(2) for master equations govern<strong>in</strong>g classical particle reaction–diffusion systems<br />
[10–12]. I will then demonstrate how the dynamic (perturbative) RG<br />
can be employed to derive the asymptotic scal<strong>in</strong>g laws <strong>in</strong> stochastic dynamical<br />
systems; to <strong>in</strong>fer the upper critical dimension d c (for dimensions d ≤ d c ,<br />
fluctuations strongly affect the universal scal<strong>in</strong>g properties); and to systematically<br />
compute the critical exponents as well as to determ<strong>in</strong>e further universal<br />
properties <strong>in</strong> various <strong>in</strong>trigu<strong>in</strong>g dynamical model systems both near and far<br />
from equilibrium. (For considerably more details, especially on the more technical<br />
aspects, the reader is referred to Ref. [13].)<br />
7.1.1 Cont<strong>in</strong>uous Phase Transitions and Critical Slow<strong>in</strong>g Down<br />
The vic<strong>in</strong>ity of a critical po<strong>in</strong>t is characterised by strong correlations and large<br />
fluctuations. The system under <strong>in</strong>vestigation is then behav<strong>in</strong>g <strong>in</strong> a highly cooperative<br />
manner, and as a consequence, the standard approximative methods of<br />
statistical mechanics, namely perturbation or cluster expansions that assume<br />
either weak <strong>in</strong>teractions or short-range correlations, fail. Upon approach<strong>in</strong>g<br />
an equilibrium cont<strong>in</strong>uous (second-order) phase transition, i.e., for |τ| ≪1,<br />
where τ =(T − T c )/T c measures the deviation from the critical temperature<br />
T c , the thermal fluctuations of the order parameter S(x) (which characterises<br />
the different thermodynamic phases, usually chosen such that the thermal<br />
average 〈S〉 = 0 vanishes <strong>in</strong> the high-temperature “disordered” phase) are, <strong>in</strong><br />
the thermodynamic limit, governed by a diverg<strong>in</strong>g length scale<br />
ξ(τ) ∼|τ| −ν . (7.1)<br />
Here, we have def<strong>in</strong>ed the correlation length via the typically exponential<br />
decay of the static cumulant or connected two-po<strong>in</strong>t correlation function<br />
C(x) =〈S(x) S(0)〉 −〈S〉 2 ∼ e −|x|/ξ ,andν denotes the correlation length<br />
critical exponent. AsT → T c , ξ →∞, which entails the absence of any characteristic<br />
length scale for the order parameter fluctuations at criticality. Hence<br />
we expect the critical correlations to follow a power law C(x) ∼|x| −(d−2+η)<br />
<strong>in</strong> d dimensions, which def<strong>in</strong>es the Fisher exponent η. The follow<strong>in</strong>g scal<strong>in</strong>g<br />
ansatz generalises this power law to T ≠ T c , but still <strong>in</strong> the vic<strong>in</strong>ity of the<br />
critical po<strong>in</strong>t,<br />
C(τ,x) =|x| −(d−2+η) ˜C± (x/ξ) , (7.2)<br />
with two dist<strong>in</strong>ct regular scal<strong>in</strong>g functions ˜C + (y) forT>T c and ˜C − (y) for<br />
T
7 Field-Theory Approaches to Nonequilibrium Dynamics 297<br />
As we will see <strong>in</strong> Subsect. 7.1.5, there are only two <strong>in</strong>dependent static<br />
critical exponents. Consequently, it must be possible to use the static scal<strong>in</strong>g<br />
hypothesis (7.2) or (7.3), along with the def<strong>in</strong>ition (7.1), to express the<br />
exponents describ<strong>in</strong>g the thermodynamic s<strong>in</strong>gularities near a second-order<br />
phase transition <strong>in</strong> terms of ν and η through scal<strong>in</strong>g laws. For example, the<br />
order parameter <strong>in</strong> the low-temperature phase (τ < 0) is expected to grow<br />
as 〈S〉 ∼ (−τ) β . Let us consider Eq. (7.2) <strong>in</strong> the limit |x| → ∞.Inorder<br />
for the |x| dependence to cancel, ˜C± (y) ∝|y| d−2+η for large |y|, and<br />
therefore C(τ,|x| → ∞) ∼ ξ −(d−2+η) ∼ |τ| ν(d−2+η) . On the other hand,<br />
C(τ,|x| →∞) →−〈S〉 2 ∼−(−τ) 2β for T < T c ; thus we identify the order<br />
parameter critical exponent through the hyperscal<strong>in</strong>g relation<br />
β = ν (d − 2+η) . (7.4)<br />
2<br />
Let us next consider the isothermal static susceptibility χ τ , which accord<strong>in</strong>g to<br />
the equilibrium fluctuation–response theorem is given <strong>in</strong> terms of the spatial<br />
<strong>in</strong>tegral of the correlation function C(τ,x): χ τ (τ) =(k B T ) −1 lim q→0 C(τ,q).<br />
But Ĉ±(p) ∼ |p| 2−η as p → 0 to ensure nons<strong>in</strong>gular behaviour, whence<br />
χ τ (τ) ∼ ξ 2−η ∼|τ| −ν(2−η) , and upon def<strong>in</strong><strong>in</strong>g the associated thermodynamic<br />
critical exponent γ via χ τ (τ) ∼|τ| −γ , we obta<strong>in</strong> the scal<strong>in</strong>g relation<br />
γ = ν (2 − η) . (7.5)<br />
The scal<strong>in</strong>g laws (7.2), (7.3) as well as scal<strong>in</strong>g relations such as (7.4) and<br />
(7.5) can be put on solid foundations by means of the RG procedure, based on<br />
an effective long-wavelength Hamiltonian H[S], a functional of S(x), that captures<br />
the essential physics of the problem, namely the relevant symmetries <strong>in</strong><br />
order parameter and real space, and the existence of a cont<strong>in</strong>uous phase transition.<br />
The probability of f<strong>in</strong>d<strong>in</strong>g a configuration S(x) at given temperature<br />
T is then given by the canonical distribution<br />
P eq [S] ∝ exp (−H[S]/k B T ) . (7.6)<br />
For example, the mathematical description of the critical phenomena for an<br />
O(n)-symmetric order parameter field S α (x), with vector <strong>in</strong>dex α =1,...,n,<br />
isbasedontheLandau–G<strong>in</strong>zburg–Wilson functional [1–6]<br />
∫<br />
H[S] = d d x ∑ [ r<br />
2 [Sα (x)] 2 + 1 2 [∇Sα (x)] 2<br />
α<br />
+ u 4! [Sα (x)] ∑ ]<br />
2 [S β (x)] 2 − h α (x) S α (x) , (7.7)<br />
β<br />
where h α (x) is the external field thermodynamically conjugate to S α (x),<br />
u>0 denotes the strength of the nonl<strong>in</strong>earity that drives the phase transformation,<br />
and r is the control parameter for the transition, i.e., r ∝ T − T 0 c ,
298 U. C. Täuber<br />
where Tc<br />
0 is the (mean-field) critical temperature. Spatial variations of the<br />
order parameter are energetically suppressed by the term ∼ [∇S α (x)] 2 ,and<br />
the correspond<strong>in</strong>g positive coefficient has been absorbed <strong>in</strong>to the fields S α .<br />
We shall, however, not pursue the static theory further here, but <strong>in</strong>stead<br />
proceed to a full dynamical description <strong>in</strong> terms of nonl<strong>in</strong>ear Langev<strong>in</strong> equations<br />
[7, 8]. We will then formulate the RG with<strong>in</strong> this dynamic framework,<br />
and there<strong>in</strong> demonstrate the emergence of scal<strong>in</strong>g laws and the computation<br />
of critical exponents <strong>in</strong> a systematic perturbative expansion with respect to<br />
the deviation ɛ = d − d c from the upper critical dimension.<br />
In order to construct the desired effective stochastic dynamics near a critical<br />
po<strong>in</strong>t, we recall that correlated region of size ξ become quite large <strong>in</strong> the<br />
vic<strong>in</strong>ity of the transition. S<strong>in</strong>ce the associated relaxation times for such clusters<br />
should grow with their extent, one would expect the characteristic time<br />
scale for the relaxation of the order parameter fluctuations to <strong>in</strong>crease as well<br />
as T → T c , namely<br />
t c (τ) ∼ ξ(τ) z ∼|τ| −zν , (7.8)<br />
which <strong>in</strong>troduces the dynamic critical exponent z that encodes the critical<br />
slow<strong>in</strong>g down at the phase transition; usually z ≥ 1. S<strong>in</strong>ce the typical relaxation<br />
rates therefore scale as ω c (τ) =1/t c (τ) ∼|τ| zν , we may utilise the<br />
static scal<strong>in</strong>g variable p = q ξ to generalise the crucial observation (7.8) and<br />
formulate a dynamic scal<strong>in</strong>g hypothesis for the wavevector-dependent dispersion<br />
relation of the order parameter fluctuations [14, 15],<br />
ω c (τ,q) =|q| z ˆω ± (q ξ) . (7.9)<br />
We can then proceed to write down dynamical scal<strong>in</strong>g laws by simply<br />
postulat<strong>in</strong>g the additional scal<strong>in</strong>g variables s = t/t c (τ) orω/ω c (τ,q). For<br />
example, as an immediate consequence we f<strong>in</strong>d for the time-dependent mean<br />
order parameter<br />
〈S(τ,t)〉 = |τ| β Ŝ(t/t c ) , (7.10)<br />
with Ŝ(s →∞) = const., but Ŝ(s) ∼ s−β/zν as s → 0 <strong>in</strong> order for the τ<br />
dependence to disappear. At the critical po<strong>in</strong>t (τ = 0), this yields the powerlaw<br />
decay 〈S(t)〉 ∼t −α , with<br />
α = β<br />
zν = 1 (d − 2+η) . (7.11)<br />
2 z<br />
Similarly, the scal<strong>in</strong>g law for the dynamic order parameter susceptibility (response<br />
function) becomes<br />
χ(τ,q,ω)=|q| −2+η ˆχ ± (q ξ,ω ξ z ) , (7.12)<br />
which constitutes the dynamical generalisation of Eq. (7.3), for χ(τ,q, 0) =<br />
(k B T ) −1 C(τ,q). Upon apply<strong>in</strong>g the fluctuation–dissipation theorem, valid <strong>in</strong><br />
thermal equilibrium, we therefrom obta<strong>in</strong> the dynamic correlation function
C(τ,q,ω)= 2k BT<br />
ω<br />
7 Field-Theory Approaches to Nonequilibrium Dynamics 299<br />
Im χ(τ,q,ω)=|q| −z−2+η Ĉ ± (q ξ,ω ξ z ) , (7.13)<br />
and for its Fourier transform <strong>in</strong> real space and time,<br />
∫ d d ∫<br />
q dω<br />
C(τ,x,t)=<br />
ei(q·x−ωt) (2π) d C(τ,q,ω)=|x| −(d−2+η) ˜C± (x/ξ, t/ξ z ) ,<br />
2π<br />
(7.14)<br />
which reduces to the static limit (7.2) if we set t =0.<br />
The critical slow<strong>in</strong>g down of the order parameter fluctuations near the<br />
critical po<strong>in</strong>t provides us with a natural separation of time scales. Assum<strong>in</strong>g<br />
(for now) that there are no other conserved variables <strong>in</strong> the system, which<br />
would constitute additional slow modes, we may thus resort to a coarse-gra<strong>in</strong>ed<br />
long-wavelength and long-time description, focus<strong>in</strong>g merely on the order parameter<br />
k<strong>in</strong>etics, while subsum<strong>in</strong>g all other “fast” degrees of freedom <strong>in</strong> random<br />
“noise” terms. This leads us to a mesoscopic Langev<strong>in</strong> equation for the slow<br />
variables S α (x,t)oftheform<br />
∂S α (x,t)<br />
= F α [S](x,t)+ζ α (x,t) . (7.15)<br />
∂t<br />
In the simplest case, the systematic force terms here just represent purely<br />
relaxational dynamics towards the equilibrium configuration [16],<br />
F α [S](x,t)=−D<br />
δH[S]<br />
δS α (x, t) , (7.16)<br />
where D represents the relaxation coefficient, and H[S] is aga<strong>in</strong> the effective<br />
Hamiltonian that governs the phase transition, e.g. given by Eq. (7.7). For<br />
the stochastic forces we may assume the most convenient form, and take them<br />
to simply represent Gaussian white noise with zero mean, 〈ζ α (x,t)〉 = 0, but<br />
with their second moment <strong>in</strong> thermal equilibrium fixed by E<strong>in</strong>ste<strong>in</strong>’s relation<br />
〈<br />
ζ α (x,t) ζ β (x ′ ,t ′ ) 〉 =2k B TDδ(x − x ′ ) δ(t − t ′ ) δ αβ . (7.17)<br />
As can be verified by means of the associated Fokker–Planck equation for the<br />
time-dependent probability distribution P[S, t], Eq. (7.17) guarantees that<br />
eventually P[S, t →∞] →P eq [S], the canonical distribution (7.6). The stochastic<br />
differential equation (7.15), with (7.16), the Hamiltonian (7.7), and<br />
the noise correlator (7.17), def<strong>in</strong>e the relaxational model A (accord<strong>in</strong>g to the<br />
classification <strong>in</strong> Ref. [7]) for a nonconserved O(n)-symmetric order parameter.<br />
If, however, the order parameter is conserved, wehavetoconsidertheassociated<br />
cont<strong>in</strong>uity equation ∂ t S α +∇·J α = 0, where typically the conserved<br />
current is given by a gradient of the field S α : J α = −D ∇S α + ...; as a consequence,<br />
the order parameter fluctuations will relax diffusively with diffusion<br />
coefficient D. The ensu<strong>in</strong>g model B [7,16] for the relaxational critical dynamics<br />
of a conserved order parameter can be obta<strong>in</strong>ed by replac<strong>in</strong>g D →−D ∇ 2 <strong>in</strong><br />
Eqs. (7.16) and (7.17). In fact, we will henceforth treat both models A and B
300 U. C. Täuber<br />
simultaneously by sett<strong>in</strong>g D → D (i∇) a , where a =0anda = 2 respectively<br />
represent the nonconserved and conserved cases. Explicitly, we thus obta<strong>in</strong><br />
∂S α (x,t)<br />
∂t<br />
= −D (i∇) a δH[S]<br />
δS α (x,t) + ζα (x,t)<br />
= −D (i∇) a[ r − ∇ 2 + u ∑<br />
[S β (x)] 2] S α (x,t)<br />
6<br />
β<br />
+D (i∇) a h α (x,t)+ζ α (x,t) , (7.18)<br />
with<br />
〈<br />
ζ α (x,t) ζ β (x ′ ,t ′ ) 〉 =2k B TD(i∇) a δ(x − x ′ ) δ(t − t ′ ) δ αβ . (7.19)<br />
Notice already that the presence or absence of a conservation law for the order<br />
parameter implies different dynamics for systems described by identical static<br />
behaviour. Before proceed<strong>in</strong>g with the analysis of the relaxational models,<br />
we remark that <strong>in</strong> general there may exist additional reversible contributions<br />
to the systematic forces F α [S], see Subsect. 7.1.6, and/or dynamical modecoupl<strong>in</strong>gs<br />
to additional conserved, slow fields, which effect further splitt<strong>in</strong>g<br />
<strong>in</strong>to several dist<strong>in</strong>ct dynamic universality classes [6, 7, 13].<br />
Let us now evaluate the dynamic response and correlation functions <strong>in</strong> the<br />
Gaussian (mean-field) approximation <strong>in</strong> the high-temperature phase. To this<br />
end, we set u = 0 and thus discard the nonl<strong>in</strong>ear terms <strong>in</strong> the Hamiltonian<br />
(7.7) as well as <strong>in</strong> Eq. (7.18). The ensu<strong>in</strong>g Langev<strong>in</strong> equation becomes l<strong>in</strong>ear <strong>in</strong><br />
the fields S α , and is therefore readily solved by means of Fourier transforms.<br />
Straightforward algebra and regroup<strong>in</strong>g some terms yields<br />
[<br />
−iω + Dq<br />
a ( r + q 2)] S α (q,ω)=Dq a h α (q,ω)+ζ α (q,ω) . (7.20)<br />
With 〈ζ α (q,ω)〉 = 0, this gives immediately<br />
χ αβ<br />
(q,ω)〉<br />
0 (q,ω)=∂〈Sα ∂h β (q,ω) ∣ = Dq a G 0 (q,ω) δ αβ , (7.21)<br />
h=0<br />
with the response propagator<br />
G 0 (q,ω)= [ −iω + Dq a (r + q 2 ) ] −1<br />
. (7.22)<br />
As is readily established by means of the residue theorem, its Fourier backtransform<br />
<strong>in</strong> time obeys causality,<br />
G 0 (q,t)=Θ(t)e −Dqa (r+q 2 ) t . (7.23)<br />
Sett<strong>in</strong>g h α = 0, and with the noise correlator (7.19) <strong>in</strong> Fourier space<br />
〈<br />
ζ α (q,ω) ζ β (q ′ ,ω ′ ) 〉 =2k B TDq a (2π) d+1 δ(q + q ′ ) δ(ω + ω ′ ) δ αβ , (7.24)
7 Field-Theory Approaches to Nonequilibrium Dynamics 301<br />
we obta<strong>in</strong> the Gaussian dynamic correlation function 〈 S α (q,ω) S β (q ′ ,ω ′ ) 〉 0 =<br />
C 0 (q,ω)(2π) d+1 δ(q + q ′ ) δ(ω + ω ′ ), where<br />
C 0 (q,ω)=<br />
2k B TDq a<br />
ω 2 +[Dq a (r + q 2 )] 2 =2k BTDq a |G 0 (q,ω)| 2 . (7.25)<br />
The fluctuation–dissipation theorem (7.13) is of course satisfied; moreover, as<br />
function of wavevector and time,<br />
C 0 (q,t)=<br />
k BT<br />
r + q 2 e−Dqa (r+q 2 ) |t| . (7.26)<br />
In the Gaussian approximation, away from criticality (r > 0, q ≠ 0) the<br />
temporal correlations for models A and B decay exponentially, with the relaxation<br />
rate ω c (r, q) = Dq 2+a (1 + r/q 2 ). Upon comparison with the dynamic<br />
scal<strong>in</strong>g hypothesis (7.9), we <strong>in</strong>fer the mean-field scal<strong>in</strong>g exponents<br />
ν 0 =1/2 andz 0 =2+a. At the critical po<strong>in</strong>t, a nonconserved order parameter<br />
relaxes diffusively (z 0 = 2) <strong>in</strong> this approximation, whereas the conserved<br />
order parameter k<strong>in</strong>etics becomes even slower, namely subdiffusive with<br />
z 0 = 4. F<strong>in</strong>ally, <strong>in</strong>vok<strong>in</strong>g Eqs. (7.12), (7.13), (7.14), or simply the static limit<br />
C 0 (q, 0) = k B T/(r + q 2 ), we f<strong>in</strong>d η 0 = 0 for the Gaussian model.<br />
The full nonl<strong>in</strong>ear Langev<strong>in</strong> equation (7.18) cannot be solved exactly. Yet<br />
a perturbation expansion with respect to the coupl<strong>in</strong>g u may be set up <strong>in</strong> a<br />
slightly cumbersome, but straightforward manner by direct iteration of the<br />
equations of motion [16, 17]. More elegantly, one may utilise a path-<strong>in</strong>tegral<br />
representation of the Langev<strong>in</strong> stochastic process [18, 19], which allows the<br />
application of all the standard tools from statistical and quantum field theory<br />
[1–6], and has the additional advantage of render<strong>in</strong>g symmetries <strong>in</strong> the<br />
problem more explicit [8, 9, 13].<br />
7.1.2 Field Theory Representation of Langev<strong>in</strong> Equations<br />
Our start<strong>in</strong>g po<strong>in</strong>t is a set of coupled Langev<strong>in</strong> equations of the form (7.15)<br />
for mesoscopic, coarse-gra<strong>in</strong>ed stochastic variables S α (x,t). For the stochastic<br />
forces, we make the simplest possible assumption of Gaussian white noise,<br />
〈ζ α (x,t)〉 =0,<br />
〈<br />
ζ α (x,t) ζ β (x ′ ,t ′ ) 〉 =2L α δ(x − x ′ ) δ(t − t ′ ) δ αβ , (7.27)<br />
where L α may represent a differential operator (such as the Laplacian ∇ 2 for<br />
conserved fields), and even a functional of S α . In the time <strong>in</strong>terval 0 ≤ t ≤ t f ,<br />
the moments (7.27) are encoded <strong>in</strong> the probability distribution<br />
[<br />
W[ζ] ∝ exp − 1 ∫ ∫ tf<br />
d d x dt ∑ ζ α (x,t) [ (L α ) −1 ζ α (x,t) ]] . (7.28)<br />
4 0 α<br />
If we now switch variables from the stochastic noise ζ α to the fields S α by<br />
means of the equations of motion (7.15), we obta<strong>in</strong>
302 U. C. Täuber<br />
W[ζ] D[ζ] =P[S] D[S] ∝ e −G[S] D[S] , (7.29)<br />
with the statistical weight determ<strong>in</strong>ed by the Onsager–Machlup functional [9]<br />
G[S] = 1 ∫ ∫<br />
d d x dt ∑ ( ( )]<br />
∂S<br />
α<br />
∂S<br />
− F α [S]<br />
)[(L α ) −1 α<br />
− F α [S] .<br />
4<br />
∂t<br />
∂t<br />
α<br />
(7.30)<br />
Note that the Jacobian for the nonl<strong>in</strong>ear variable transformation {ζ α }→<br />
{S α } has been omitted here. In fact, the above procedure is properly def<strong>in</strong>ed<br />
through appropriately discretis<strong>in</strong>g time. If a forward (Itô) discretisation is applied,<br />
then <strong>in</strong>deed the associated functional determ<strong>in</strong>ant is a mere constant<br />
that can be absorbed <strong>in</strong> the functional measure. The functional (7.30) already<br />
represents a desired field theory action. S<strong>in</strong>ce the probability distribution for<br />
the stochastic forces should be normalised, ∫ D[ζ] W [ζ] = 1, the associated<br />
“partition function” is unity, and carries no physical <strong>in</strong>formation (as opposed<br />
to static statistical field theory, where it determ<strong>in</strong>es the free energy and hence<br />
the entire thermodynamics). The Onsager–Machlup representation is however<br />
plagued by technical problems: Eq. (7.30) conta<strong>in</strong>s (L α ) −1 , which for<br />
conserved variables entails the <strong>in</strong>verse Laplacian operator, i.e., a Green function<br />
<strong>in</strong> real space or the s<strong>in</strong>gular factor 1/q 2 <strong>in</strong> Fourier space; moreover the<br />
nonl<strong>in</strong>earities <strong>in</strong> F α [S] appear quadratically. Hence it is desirable to l<strong>in</strong>earise<br />
the action (7.30) by means of a Hubbard–Stratonovich transformation [9].<br />
We shall follow an alternative, more general route that completely avoids<br />
the appearance of the <strong>in</strong>verse operators (L α ) −1 <strong>in</strong> <strong>in</strong>termediate steps. Our<br />
goal is to average over noise “histories” for observables A[S] that need to be<br />
expressible <strong>in</strong> terms of the stochastic fields S α : 〈A[S]〉 ζ ∝ ∫ D[ζ] A[S(ζ)] W [ζ].<br />
For this purpose, we employ the identity<br />
∫<br />
1= D[S] ∏ ∏<br />
( ∂S α )<br />
(x,t)<br />
δ<br />
− F α [S](x,t) − ζ α (x,t)<br />
∂t<br />
α (x,t)<br />
∫ ∫ [ ∫ ∫<br />
= D[i ˜S] D[S]exp − d d x dt ∑ ( )]<br />
∂S<br />
˜S α α<br />
− F α [S] − ζ α , (7.31)<br />
∂t<br />
α<br />
where the first l<strong>in</strong>e constitutes a rather <strong>in</strong>volved representation of the unity (<strong>in</strong><br />
a somewhat symbolic notation; aga<strong>in</strong> proper discretisation should be <strong>in</strong>voked<br />
here), and the second l<strong>in</strong>e utilises the Fourier representation of the (functional)<br />
delta distribution by means of the purely imag<strong>in</strong>ary auxiliary fields ˜S (and<br />
factors 2π have been absorbed <strong>in</strong> its functional measure).<br />
Insert<strong>in</strong>g (7.31) and the probability distribution (7.28) <strong>in</strong>to the desired<br />
stochastic noise average, we arrive at<br />
∫ ∫ [ ∫ ∫<br />
〈A[S]〉 ζ ∝ D[i ˜S] D[S]exp − d d x dt ∑ ( )]<br />
∂S<br />
˜S α α<br />
− F α [S] A[S]<br />
∂t<br />
α<br />
∫ ( ∫ ∫<br />
× D[ζ]exp − d d x dt ∑ [ ])<br />
1<br />
4 ζα (L α ) −1 ζ α − ˜S α ζ α . (7.32)<br />
α
7 Field-Theory Approaches to Nonequilibrium Dynamics 303<br />
We may now evaluate the Gaussian <strong>in</strong>tegrals over the noise ζ α , which yields<br />
∫<br />
∫<br />
〈A[S]〉 ζ = D[S] A[S] P[S] , P[S] ∝ D[i ˜S]e −A[ ˜S,S] , (7.33)<br />
with the statistical weight now governed by the Janssen–De Dom<strong>in</strong>icis “response”<br />
functional [9, 18, 19]<br />
∫ ∫<br />
A[ ˜S,S]=<br />
tf<br />
d d x dt ∑ [ ( )<br />
]<br />
∂S<br />
˜S α α<br />
− F α [S] −<br />
0<br />
∂t<br />
˜S α L α ˜Sα . (7.34)<br />
α<br />
Once aga<strong>in</strong>, we have omitted the functional determ<strong>in</strong>ant from the variable<br />
change {ζ α }→{S α }, and normalisation implies ∫ D[i ˜S] ∫ D[S]e −A[ ˜S,S] =1.<br />
The first term <strong>in</strong> the action (7.34) encodes the temporal evolution accord<strong>in</strong>g<br />
to the systematic terms <strong>in</strong> the Langev<strong>in</strong> equations (7.15), whereas the second<br />
term specifies the noise correlations (7.27). S<strong>in</strong>ce the auxiliary variables ˜S α ,<br />
often termed Mart<strong>in</strong>–Siggia–Rose response fields [20], appear only quadratically<br />
here, they may be elim<strong>in</strong>ated via complet<strong>in</strong>g the squares and Gaussian<br />
<strong>in</strong>tegrations; thereby one recovers the Onsager–Machlup functional (7.30).<br />
The Janssen–De Dom<strong>in</strong>icis functional (7.34) takes the form of a (d + 1)-<br />
dimensional statistical field theory with two <strong>in</strong>dependent sets of fields S α and<br />
˜S α . We may thus br<strong>in</strong>g the established mach<strong>in</strong>ery of statistical and quantum<br />
field theory [1–6] to bear here; it should however be noted that the response<br />
functional formalism for stochastic Langev<strong>in</strong> dynamics <strong>in</strong>corporates causality<br />
<strong>in</strong> a nontrivial manner, which leads to important dist<strong>in</strong>ctions [8].<br />
Let us specify the Janssen–De Dom<strong>in</strong>icis functional for the purely relaxational<br />
models A and B [16, 17], see Eqs. (7.18) and (7.19), splitt<strong>in</strong>g it <strong>in</strong>to<br />
the Gaussian and anharmonic parts A = A 0 + A <strong>in</strong>t [9], which read<br />
∫ ∫<br />
A 0 [ ˜S,S] = d d x dt ∑ ( [ ]<br />
∂<br />
˜S α ∂t + D (i∇)a (r − ∇ 2 ) S α<br />
α<br />
)<br />
−D ˜S α (i∇) a ˜Sα − D ˜S α (i∇) a h α , (7.35)<br />
A <strong>in</strong>t [ ˜S,S] =D u ∫ ∫<br />
d d x dt ∑ ˜S α (i∇) a S α S β S β . (7.36)<br />
6<br />
α,β<br />
S<strong>in</strong>ce we are <strong>in</strong>terested <strong>in</strong> the vic<strong>in</strong>ity of the critical po<strong>in</strong>t T ≈ T c ,wehave<br />
absorbed the constant k B T c <strong>in</strong>to the fields. The prescription (7.33) tells us how<br />
to compute time-dependent correlation functions 〈 S α (x,t) S β (x ′ ,t ′ ) 〉 .Us<strong>in</strong>g<br />
Eq. (7.35), the dynamic order parameter susceptibility follows from<br />
χ αβ (x − x ′ ,t− t ′ )= δ〈Sα (x,t)〉<br />
δh β (x ′ ,t ′ )<br />
∣ = D<br />
h=0<br />
〈<br />
〉<br />
S α (x,t)(i∇) a ˜Sβ (x ′ ,t ′ )<br />
(7.37)<br />
for the simple relaxational models (only), the response function is just given by<br />
a correlator that <strong>in</strong>volves an auxiliary variable, which expla<strong>in</strong>s why the ˜S α are<br />
;
304 U. C. Täuber<br />
referred to as “response” fields. In equilibrium, one may employ the Onsager–<br />
Machlup functional (7.30) to derive the fluctuation–dissipation theorem [9]<br />
χ αβ (x − x ′ ,t− t ′ )=Θ(t − t ′ ) ∂<br />
∂t ′ 〈<br />
S α (x,t) S β (x ′ ,t ′ ) 〉 , (7.38)<br />
which is equivalent to Eq. (7.13) <strong>in</strong> Fourier space.<br />
In order to access arbitrary correlators, we def<strong>in</strong>e the generat<strong>in</strong>g functional<br />
〈 ∫ ∫<br />
Z[˜j,j]= exp d d x dt ∑ (˜j α ˜Sα + j α S α)〉 , (7.39)<br />
α<br />
wherefrom the correlation functions follow via functional derivatives,<br />
〈 ∏<br />
〉<br />
S αi ˜S αj = ∏ δ δ<br />
Z[˜j,j] ∣ , (7.40)<br />
δj αi δ˜j<br />
ij<br />
ij<br />
αj ∣˜j=0=j<br />
and the cumulants or connected correlation functions via<br />
〈 ∏<br />
〉<br />
S αi ˜S αj = ∏ δ δ<br />
ln Z[˜j,j] ∣<br />
ij<br />
c<br />
δj αi δ˜j<br />
ij<br />
αj<br />
∣˜j=0=j<br />
. (7.41)<br />
In the harmonic approximation, sett<strong>in</strong>g u =0,Z[˜j,j] can be evaluated explicitly<br />
(most directly <strong>in</strong> Fourier space) by means of Gaussian <strong>in</strong>tegration [9,13];<br />
one thereby recovers (with k B T = 1) the Gaussian response propagator (7.22)<br />
and two-po<strong>in</strong>t 〈 correlation function (7.25). Moreover, as a consequence of<br />
causality, ˜Sα (q,ω) ˜S β (q ′ ,ω )〉<br />
′ =0.<br />
7.1.3 Outl<strong>in</strong>e of Dynamic Perturbation Theory<br />
0<br />
S<strong>in</strong>ce we cannot evaluate correlation functions with the nonl<strong>in</strong>ear action (7.36)<br />
exactly, we resort to a perturbational treatment, assum<strong>in</strong>g, for the time be<strong>in</strong>g,<br />
a small coupl<strong>in</strong>g strength u. Theperturbation expansion with respect to u is<br />
constructed by rewrit<strong>in</strong>g the desired correlation functions <strong>in</strong> terms of averages<br />
with respect to the Gaussian action (7.35), henceforth <strong>in</strong>dicated with <strong>in</strong>dex<br />
’0’, and then expand<strong>in</strong>g the exponential of −A <strong>in</strong>t ,<br />
〈 ∏<br />
ij<br />
S αi ˜S αj 〉<br />
=<br />
〈 ∏ ˜S ij Sαi αj −A<strong>in</strong>t[ ˜S,S]〉<br />
e<br />
〈 ∏<br />
=<br />
ij<br />
〈 〉<br />
e−A<strong>in</strong>t[ ˜S,S]<br />
0<br />
S αi ˜S ∑ ∞ αj<br />
l=0<br />
1<br />
l!<br />
0<br />
(<br />
−A <strong>in</strong>t [ ˜S,S]<br />
) l<br />
〉<br />
0<br />
. (7.42)<br />
The rema<strong>in</strong><strong>in</strong>g Gaussian averages, a series of polynomials <strong>in</strong> the fields S α<br />
and ˜S α , can be evaluated by means of Wick’s theorem, here an immediate
7 Field-Theory Approaches to Nonequilibrium Dynamics 305<br />
consequence of the Gaussian statistical weight, which states that all such averages<br />
can be written as a sum over all possible factorisations <strong>in</strong>to Gaussian<br />
two-po<strong>in</strong>t functions 〈S α ˜Sβ 〉 0 , i.e., essentially the response propagator G 0 ,<br />
Eq. (7.22), and 〈S α S β 〉 0 , the Gaussian correlation function C 0 , Eq. (7.25).<br />
Recall that the denom<strong>in</strong>ator <strong>in</strong> Eq. (7.42) is exactly unity as a consequence of<br />
normalisation; alternatively, this result follows from causality <strong>in</strong> conjunction<br />
with our forward descretisation prescription, which implies that we should<br />
identify Θ(0) = 0. (We remark that had we chosen another temporal discretisation<br />
rule, any apparent contributions from the denom<strong>in</strong>ator would be<br />
precisely cancelled by the <strong>in</strong> this case nonvanish<strong>in</strong>g functional Jacobian from<br />
the variable transformation {ζ α }→{S α }.) At any rate, our stochastic field<br />
theory conta<strong>in</strong>s no “vacuum” contributions.<br />
The many terms <strong>in</strong> the perturbation expansion (7.42) are most lucidly<br />
organised <strong>in</strong> a graphical representation, us<strong>in</strong>g Feynman diagrams with the<br />
basic elements depicted <strong>in</strong> Fig. 7.1. We represent the response propagator<br />
(7.22) by a directed l<strong>in</strong>e (here conventionally from right to left), which encodes<br />
its causal nature; the noise by a two-po<strong>in</strong>t “source” vertex, and the anharmonic<br />
term <strong>in</strong> Eq. (7.36) as a four-po<strong>in</strong>t vertex. In the diagrams represent<strong>in</strong>g the<br />
different terms <strong>in</strong> the perturbation series, these vertices serve as l<strong>in</strong>ks for the<br />
propagator l<strong>in</strong>es, with the fields S α be<strong>in</strong>g encoded as the “<strong>in</strong>com<strong>in</strong>g”, and the<br />
˜S α as the “outgo<strong>in</strong>g” components of the l<strong>in</strong>es. In Fourier space, translational<br />
<strong>in</strong>variance <strong>in</strong> space and time implies wavevector and frequency conservation<br />
at each vertex, see Fig. 7.2 below. An alternative, equivalent representation<br />
uses both the response and correlation propagators as <strong>in</strong>dependent elements,<br />
the latter depicted as undirected l<strong>in</strong>e, thereby dispos<strong>in</strong>g of the noise vertex,<br />
and reta<strong>in</strong><strong>in</strong>g the nonl<strong>in</strong>earity <strong>in</strong> Fig. 7.1(c) as sole vertex.<br />
Follow<strong>in</strong>g standard field theory procedures [1–5], one establishes that the<br />
perturbation series for the cumulants (7.41) is given <strong>in</strong> terms of connected<br />
Feynman graphs only (for a detailed exposition of this and the follow<strong>in</strong>g results,<br />
see Ref. [13]). An additional helpful reduction <strong>in</strong> the number of diagrams<br />
to be considered arises when one considers the vertex functions, which generalise<br />
the self-energy contributions Σ(q,ω)<strong>in</strong>theDyson equation for the<br />
response propagator, G(q,ω) −1 = Dq a χ(q,ω) −1 = G 0 (q,ω) −1 − Σ(q,ω).<br />
(a)<br />
α<br />
q,ω<br />
β<br />
=<br />
1<br />
i ω+ D q<br />
a<br />
( r + q2)<br />
δ αβ<br />
(b)<br />
α<br />
β<br />
q<br />
-q<br />
=<br />
2 D q a δ αβ<br />
(c)<br />
α<br />
q<br />
α<br />
β<br />
β<br />
=<br />
D q a<br />
u<br />
6<br />
Fig. 7.1. Elements of dynamic perturbation theory for the O(n)-symmetric relaxational<br />
models: (a) response propagator; (b) noisevertex;(c) anharmonic vertex
306 U. C. Täuber<br />
To this end, we def<strong>in</strong>e the fields ˜Φ α = δ ln Z/δ˜j α and Φ α = δ ln Z/δj α ,and<br />
<strong>in</strong>troduce the new generat<strong>in</strong>g functional<br />
∫ ∫<br />
Γ [ ˜Φ, Φ] =− ln Z[˜j,j]+ d d x dt ∑ (˜j α ˜Φα + j α Φ α) , (7.43)<br />
α<br />
wherefrom the vertex functions are obta<strong>in</strong>ed via the functional derivatives<br />
Γ (Ñ,N)<br />
{α i};{α j} = Ñ<br />
∏<br />
i<br />
N<br />
δ<br />
αi<br />
δ ˜Φ<br />
∏<br />
δ<br />
δΦ Γ [ ˜Φ, Φ] ∣<br />
αj<br />
j<br />
∣˜j=0=j<br />
. (7.44)<br />
Diagrammatically, these quantities turn out to be represented by the possible<br />
sets of one-particle (1PI) irreducible Feynman graphs with N <strong>in</strong>com<strong>in</strong>g and<br />
Ñ outgo<strong>in</strong>g “amputated” legs; i.e., these diagrams do not split <strong>in</strong>to allowed<br />
subgraphs by simply cutt<strong>in</strong>g any s<strong>in</strong>gle propagator l<strong>in</strong>e. For example, for the<br />
two-po<strong>in</strong>t functions a direct calculation yields the relations<br />
Γ (1,1) (q,ω)=Dq a χ(−q, −ω) −1 = G 0 (−q, −ω) −1 − Σ(−q, −ω) ,(7.45)<br />
Γ (2,0) (q,ω)=−<br />
C(q,ω) qa<br />
= −2D Im Γ (1,1) (q,ω) , (7.46)<br />
|G(q,ω)|<br />
2<br />
ω<br />
where the second equation for Γ (2,0) follows from the fluctuation–dissipation<br />
theorem (7.13). Note that Γ (0,2) (q,ω) = 0 vanishes because of causality.<br />
The perturbation series can then be organised graphically as an expansion<br />
<strong>in</strong> successive orders with respect to the number of closed propagator loops.<br />
As an example, Fig. 7.2 depicts the one-loop contributions for the vertex<br />
functions Γ (1,1) and Γ (1,3) <strong>in</strong> the time doma<strong>in</strong> with all required labels. One<br />
may formulate general Feynman rules for the construction of the diagrams and<br />
their translation <strong>in</strong>to mathematical expressions for the lth order contribution<br />
to the vertex function Γ (Ñ,N) :<br />
1. Draw all topologically different, connected one-particle irreducible graphs<br />
with Ñ outgo<strong>in</strong>g and N <strong>in</strong>com<strong>in</strong>g l<strong>in</strong>es connect<strong>in</strong>g l relaxation vertices ∝<br />
u. Donot allow closed response loops (s<strong>in</strong>ce <strong>in</strong> the Itô calculus Θ(0) = 0).<br />
2. Attach wavevectors q i , frequencies ω i or times t i , and component <strong>in</strong>dices<br />
α i to all directed l<strong>in</strong>es, obey<strong>in</strong>g “momentum (and energy)” conservation<br />
at each vertex.<br />
(a)<br />
α<br />
k<br />
t<br />
t´<br />
β -k<br />
α<br />
(b)<br />
α<br />
0<br />
q-k<br />
β<br />
t<br />
k<br />
α<br />
α<br />
α<br />
-k<br />
t´<br />
Fig. 7.2. One-loop diagrams for (a) Γ (1,1) and (b) Γ (1,3) <strong>in</strong> the time doma<strong>in</strong>
7 Field-Theory Approaches to Nonequilibrium Dynamics 307<br />
3. Each directed l<strong>in</strong>e corresponds to a response propagator G 0 (−q, −ω) or<br />
G 0 (q,t i −t j ) <strong>in</strong> the frequency and time doma<strong>in</strong>, respectively, the two-po<strong>in</strong>t<br />
vertex to the noise strength 2D q a , and the four-po<strong>in</strong>t relaxation vertex<br />
to −D q a u/6. Closed loops imply <strong>in</strong>tegrals over the <strong>in</strong>ternal wavevectors<br />
and frequencies or times, subject to causality constra<strong>in</strong>ts, as well as sums<br />
over the <strong>in</strong>ternal vector <strong>in</strong>dices. Apply the residue theorem to evaluate<br />
frequency <strong>in</strong>tegrals.<br />
4. Multiply with −1 and the comb<strong>in</strong>atorial factor count<strong>in</strong>g all possible ways<br />
of connect<strong>in</strong>g the propagators, l relaxation vertices, and k two-po<strong>in</strong>t vertices<br />
lead<strong>in</strong>g to topologically identical graphs, <strong>in</strong>clud<strong>in</strong>g a factor 1/l! k!<br />
orig<strong>in</strong>at<strong>in</strong>g <strong>in</strong> the expansion of exp(−A <strong>in</strong>t [ ˜S,S]).<br />
For later use, we provide the explicit results for the two-po<strong>in</strong>t vertex functions<br />
to two-loop order. After some algebra, the three diagrams <strong>in</strong> Fig. 7.3<br />
give<br />
[<br />
Γ (1,1) (q,ω)=iω + Dq a r + q 2 + n +2<br />
6<br />
( ) 2 ∫<br />
n +2<br />
− u<br />
6<br />
k<br />
− n +2<br />
×<br />
18 u2 ∫k<br />
(<br />
1 −<br />
∫<br />
1<br />
r + k 2<br />
∫<br />
1<br />
r + k 2<br />
∫<br />
u<br />
k<br />
k ′ 1<br />
1<br />
r + k 2<br />
1<br />
k ′ (r + k ′2 ) 2<br />
1<br />
r + k ′ 2<br />
r +(q − k − k ′ )<br />
)]<br />
2<br />
, (7.47)<br />
iω<br />
iω + ∆(k)+∆(k ′ )+∆(q − k − k ′ )<br />
where we have separated out the dynamic part <strong>in</strong> the last l<strong>in</strong>e, and <strong>in</strong>troduced<br />
the abbreviations ∆(q) =Dq a (r + q 2 )and ∫ k = ∫ d d k/(2π) d [13]. For the<br />
noise vertex, Fig. 7.4(a) yields [13]<br />
[<br />
Γ (2,0) (q,ω)=−2Dq a 1+Dq a n +2<br />
18<br />
∫k<br />
u2<br />
1<br />
×<br />
r +(q − k − k ′ ) 2 Re 1<br />
∫<br />
1<br />
r + k 2<br />
k ′ 1<br />
r + k ′ 2<br />
]<br />
iω + ∆(k)+∆(k ′ )+∆(q − k − k ′ ; (7.48)<br />
)<br />
notice that for model B, as a consequence of the conservation law for the<br />
order parameter and ensu<strong>in</strong>g wavevector dependence of the nonl<strong>in</strong>ear vertex,<br />
+ + + ...<br />
Fig. 7.3. One-particle irreducible diagrams for Γ (1,1) (q,ω) to second order <strong>in</strong> u
308 U. C. Täuber<br />
(a)<br />
(b)<br />
Fig. 7.4. (a) Two-loop diagram for Γ (2,0) (q,ω); (b) one-loop graph for Γ (1,3)<br />
see Fig. 7.1(c), to all orders <strong>in</strong> the perturbation expansion<br />
a =2 : Γ (1,1) ∂<br />
(q =0,ω)=iω,<br />
∂q 2 Γ (2,0) (q,ω)<br />
∣ = −2D . (7.49)<br />
q=0<br />
At last, with the shorthand notation k =(q,ω), the analytical expression<br />
correspond<strong>in</strong>g to the graph <strong>in</strong> Fig. 7.4(b) for the four-po<strong>in</strong>t vertex function<br />
at symmetrically chosen external wavevector labels reads<br />
( ) a [ 3<br />
Γ (1,3) (−3k/2; {k/2}) =D<br />
2 q u<br />
∫<br />
(<br />
1 1<br />
×<br />
k r + k 2 r +(q − k) 2 1 −<br />
7.1.4 Renormalisation<br />
1 − n +8<br />
6<br />
u<br />
iω<br />
iω + ∆(k)+∆(q − k)<br />
)]<br />
. (7.50)<br />
Consider a typical loop <strong>in</strong>tegral, say the correction <strong>in</strong> Eq. (7.50) to the fourpo<strong>in</strong>t<br />
vertex function Γ (1,3) at zero external frequency and momentum, whose<br />
“bare” value, without any fluctuation contributions, is u. In dimensions d
7 Field-Theory Approaches to Nonequilibrium Dynamics 309<br />
Conversely, for dimensions d ≥ 4, the <strong>in</strong>tegral <strong>in</strong> (7.51) develops ultraviolet<br />
(UV) divergences as the upper <strong>in</strong>tegral boundary is sent to <strong>in</strong>f<strong>in</strong>ity (k = |k|),<br />
∫ Λ<br />
k d−1 { }<br />
ln(Λ<br />
(r + k 2 ) 2 dk ∼ 2 /r) d =4<br />
Λ d−4 →∞ as Λ →∞. (7.52)<br />
d>4<br />
0<br />
In lattice models, there is a f<strong>in</strong>ite wavevector cutoff, namely the Brillou<strong>in</strong><br />
zone boundary, Λ ∼ (2π/a 0 ) d for a hypercubic lattice with lattice constant<br />
a 0 , whence physically these UV problems do not emerge. Yet we shall see that<br />
a formal treatment of these unphysical UV divergences will allow us to <strong>in</strong>fer<br />
the correct power laws for the physical IR s<strong>in</strong>gularities associated with the<br />
critical po<strong>in</strong>t. The borderl<strong>in</strong>e dimension that separates the IR and UV s<strong>in</strong>gular<br />
regimes is referred to as upper critical dimension d c ; here d c = 4. Note that<br />
at d c , UV and IR s<strong>in</strong>gularities are <strong>in</strong>timately connected and appear <strong>in</strong> the<br />
form of logarithmic divergences, see Eq. (7.52). The situation is summarised<br />
<strong>in</strong> Table 7.1, where we have also stated that models with cont<strong>in</strong>uous order<br />
parameter symmetry, such as the Hamiltonian (7.7) with n ≥ 2, do not allow<br />
long-range order <strong>in</strong> dimensions d ≤ d lc = 2 (Merm<strong>in</strong>–Wagner–Hohenberg<br />
theorem [21–23]). Here, d lc is called the lower critical dimension; for the Is<strong>in</strong>g<br />
model represented by Eq. (7.7) with n =1,ofcoursed lc =1.<br />
Table 7.1. Mathematical and physical dist<strong>in</strong>ctions of the regimes dd c, for the O(n)-symmetric models A and B (or static Φ 4 field theory)<br />
Dimension Perturbation Model A / B or Critical<br />
Interval Series Φ 4 Field Theory Behaviour<br />
d ≤ d lc = 2 IR-s<strong>in</strong>gular ill-def<strong>in</strong>ed no long-range<br />
UV-convergent u relevant order (n ≥ 2)<br />
2
310 U. C. Täuber<br />
phase transition; but u becomes irrelevant for d>4: one then expects meanfield<br />
(Gaussian) critical exponents. At the upper critical dimension d c =4,<br />
the nonl<strong>in</strong>ear coupl<strong>in</strong>g u is marg<strong>in</strong>ally relevant: this will <strong>in</strong>duce logarithmic<br />
corrections to the mean-field scal<strong>in</strong>g laws, see Table 7.1.<br />
It is obviously not a simple task to treat the IR-s<strong>in</strong>gular perturbation<br />
expansion <strong>in</strong> a mean<strong>in</strong>gful, well-def<strong>in</strong>ed manner, and thus allow nonanalytic<br />
modifications of the critical power laws (note that mean-field scal<strong>in</strong>g is completely<br />
determ<strong>in</strong>ed by dimensional analysis or power count<strong>in</strong>g). The key of<br />
the success of the RG approach is to focus on the very specific symmetry<br />
that emerges near critical po<strong>in</strong>ts, namely scale <strong>in</strong>variance. There are several<br />
(largely equivalent) versions of the RG method; we shall here formulate<br />
and employ the field-theoretic variant [1–6, 8, 13]. In order to proceed,<br />
it is convenient to evaluate the loop <strong>in</strong>tegrals <strong>in</strong> momentum space by means<br />
of dimensional regularisation, whereby one assigns f<strong>in</strong>ite values even to UVdivergent<br />
expressions, namely the analytically cont<strong>in</strong>ued values from the UVf<strong>in</strong>ite<br />
range. For example, even for non<strong>in</strong>teger dimensions d and σ, weset<br />
∫<br />
d d k<br />
(2π) d<br />
k 2σ Γ (σ + d/2) Γ (s − σ − d/2)<br />
(τ + k 2 ) s =<br />
2 d π d/2 Γ (d/2) Γ (s)<br />
The renormalisation program then consists of the follow<strong>in</strong>g steps:<br />
τ σ−s+d/2 . (7.53)<br />
1. We aim to carefully keep track of formal, unphysical UV divergences. In<br />
dimensionally regularised <strong>in</strong>tegrals (7.53), these appear as poles <strong>in</strong> ɛ =<br />
d c − d; their residues characterise the asymptotic UV behaviour of the field<br />
theory under consideration.<br />
2. Therefrom we may <strong>in</strong>fer the (UV) scal<strong>in</strong>g properties of the control parameters<br />
of the model under a RG transformation, namely essentially a change<br />
of the momentum scale µ, while keep<strong>in</strong>g the form of the action <strong>in</strong>variant.<br />
This will allow us to def<strong>in</strong>e suitable runn<strong>in</strong>g coupl<strong>in</strong>gs.<br />
3. We seek fixed po<strong>in</strong>ts <strong>in</strong> parameter space where certa<strong>in</strong> marg<strong>in</strong>al coupl<strong>in</strong>gs<br />
(u here) do not change anymore under RG transformations. This describes<br />
a scale-<strong>in</strong>variant regime for the model under consideration, where the UV<br />
and IR scal<strong>in</strong>g properties become <strong>in</strong>timately l<strong>in</strong>ked. Study<strong>in</strong>g the parameter<br />
flows near a stable RG fixed po<strong>in</strong>t then allows us to extract the<br />
asymptotic IR power laws.<br />
As a prelim<strong>in</strong>ary step, we need to take <strong>in</strong>to account that the fluctuations<br />
will also shift the critical po<strong>in</strong>t downwards from the mean-field phase transition<br />
temperature T 0 c ; i.e., we expect the transition to occur at T c
7 Field-Theory Approaches to Nonequilibrium Dynamics 311<br />
Notice that this quantity depends on microscopic details (the lattice structure<br />
enters the cutoff Λ) and is thus not universal; moreover it diverges for d ≥ 2<br />
(quadratically near d c =4)asΛ →∞.Wenextuser = τ + r c to write<br />
physical quantities as functions of the true distance τ from the critical po<strong>in</strong>t,<br />
which technically amounts to an additive renormalisation; e.g., the dynamic<br />
response function becomes to one-loop order<br />
χ(q,ω) −1 = − iω<br />
Dq a + q2 + τ<br />
[<br />
1 − n +2 u<br />
6<br />
∫<br />
k<br />
]<br />
1<br />
k 2 (τ + k 2 + O(u 2 ) . (7.55)<br />
)<br />
The rema<strong>in</strong><strong>in</strong>g loop <strong>in</strong>tegral is UV-s<strong>in</strong>gular <strong>in</strong> dimensions d ≥ d c =4.<br />
We may now formally absorb the rema<strong>in</strong><strong>in</strong>g UV divergences <strong>in</strong>to renormalised<br />
fields and parameters, a procedure called multiplicative renormalisation.<br />
For the renormalised fields, we use the convention<br />
SR α = Z 1/2<br />
S<br />
S α , ˜Sα R = Z 1/2 ˜S α , (7.56)<br />
where we have exploited the O(n) rotational symmetry <strong>in</strong> us<strong>in</strong>g identical<br />
renormalisation constants (Z factors) for each component. The renormalised<br />
cumulants with N order parameter fields S α and Ñ response fields ˜S α naturally<br />
<strong>in</strong>volve the product Z N/2<br />
S<br />
Γ (Ñ,N)<br />
R<br />
ZÑ/2 , whence<br />
˜S<br />
˜S<br />
= Z −Ñ/2 Z −N/2<br />
˜S S<br />
Γ (Ñ,N) . (7.57)<br />
In a similar manner, we relate the “bare” parameters of the theory via Z<br />
factors to their renormalised counterparts, which we furthermore render dimensionless<br />
through appropriate momentum scale factors,<br />
D R = Z D D, τ R = Z τ τµ −2 , u R = Z u uA d µ d−4 , (7.58)<br />
where we have separated out the factor A d = Γ (3 − d/2)/2 d−1 π d/2 for convenience.<br />
In the m<strong>in</strong>imal subtraction scheme, the Z factors conta<strong>in</strong> only the<br />
UV-s<strong>in</strong>gular terms, which <strong>in</strong> dimensional regularisation appear as poles at<br />
ɛ = 0, and their residues, evaluated at d = d c .<br />
These renormalisation constants are not all <strong>in</strong>dependent, however; s<strong>in</strong>ce<br />
the equilibrium fluctuation–dissipation theorem (7.38) or (7.13) must hold <strong>in</strong><br />
the renormalised theory as well, we <strong>in</strong>fer that necessarily<br />
and consequently from Eq. (7.45)<br />
Z D = ( Z S /Z ˜S) 1/2<br />
, (7.59)<br />
χ R = Z S χ. (7.60)<br />
Moreover, for model B with conserved order parameter Eq. (7.49) implies that<br />
to all orders <strong>in</strong> the perturbation expansion
312 U. C. Täuber<br />
a =2 : Z ˜S<br />
Z S =1, Z D = Z S . (7.61)<br />
For the follow<strong>in</strong>g, it is crucial that the theory is renormalisable, i.e., a f<strong>in</strong>ite<br />
number of reparametrisations suffice to formally rid it of all UV divergences.<br />
Indeed, for the relaxational models A and B, and the static G<strong>in</strong>zburg–Landau–<br />
Wilson Hamiltonian (7.7), all higher vertex function beyond the four-po<strong>in</strong>t<br />
function are UV-convergent near d c , and there are only the three <strong>in</strong>dependent<br />
static renormalisation factors Z S , Z τ ,andZ u , and <strong>in</strong> addition Z D for nonconserved<br />
order parameter dynamics. As we shall see, these directly translate<br />
<strong>in</strong>to the two <strong>in</strong>dependent static critical exponents and the unrelated dynamic<br />
scal<strong>in</strong>g exponent z for model A; for model B with conserved order parameter,<br />
Eq. (7.61) will yield a scal<strong>in</strong>g relation between z and η.<br />
In order to explicitly determ<strong>in</strong>e the renormalisation constants, we need<br />
to ensure that we stay away from the IR-s<strong>in</strong>gular regime. This is guaranteed<br />
by select<strong>in</strong>g as normalisation po<strong>in</strong>t either τ R = 1 (i.e., Z τ τ = µ 2 )orq = µ.<br />
Inevitably therefore, the renormalised theory depends on the correspond<strong>in</strong>g<br />
arbitrary momentum scale µ. S<strong>in</strong>ce there are no fluctuation contributions to<br />
order u to either ∂Γ (1,1) (q, 0)/∂q 2 or ∂Γ (1,1) (0,ω)/∂ω (at τ R = 1), we f<strong>in</strong>d<br />
Z S =1andZ D = 1 with<strong>in</strong> the one-loop approximation. Expressions (7.55)<br />
and (7.50) then yield with the formula (7.53)<br />
Z τ =1− n +2<br />
6<br />
u R<br />
ɛ<br />
, Z u =1− n +8<br />
6<br />
u R<br />
ɛ<br />
. (7.62)<br />
To two-loop order, we may <strong>in</strong>fer the field renormalisation Z S from the static<br />
susceptibility as the s<strong>in</strong>gular contributions to ∂χ R (q, 0)/∂q 2 | q=0 ,andZ D for<br />
model A, through a somewhat lengthy calculation [13], from either Γ (2,0)<br />
R<br />
(0, 0)<br />
or Γ (1,1)<br />
R<br />
(0,ω), with the results<br />
Z S =1+ n +2<br />
144<br />
u 2 R<br />
ɛ<br />
, a =0 : Z D =1− n +2<br />
144<br />
7.1.5 Scal<strong>in</strong>g Laws and Critical Exponents<br />
(<br />
6ln 4 3 − 1 ) u<br />
2<br />
R<br />
ɛ<br />
. (7.63)<br />
We now wish to related the renormalised vertex functions at different <strong>in</strong>verse<br />
length scales µ. This is accomplished by simply recall<strong>in</strong>g that the unrenormalised<br />
vertex functions obviously do not depend on µ,<br />
0=µ d<br />
dµ Γ (Ñ,N) (D, τ, u) =µ d [<br />
dµ<br />
ZÑ/2 ˜S<br />
]<br />
Z N/2<br />
S<br />
Γ (Ñ,N)<br />
R<br />
(µ, D R ,τ R ,u R ) . (7.64)<br />
In the second step, the bare quantities have been replaced with their renormalised<br />
counterparts. The <strong>in</strong>nocuous statement (7.64) then implies a very<br />
nontrivial partial differential equation for the renormalised vertex functions,<br />
the desired renormalisation group equation,
7 Field-Theory Approaches to Nonequilibrium Dynamics 313<br />
[<br />
µ ∂<br />
∂µ + Ñγ˜S + Nγ S<br />
∂<br />
+ γ D D R + γ τ τ R<br />
2<br />
∂D R<br />
∂<br />
∂τ R<br />
+ β u<br />
]<br />
∂<br />
∂u R<br />
× Γ (Ñ,N)<br />
R<br />
(µ, D R ,τ R ,u R )=0. (7.65)<br />
Here we have def<strong>in</strong>ed Wilson’s flow functions (the <strong>in</strong>dex “0” <strong>in</strong>dicates that<br />
the derivatives with respect to µ are to be taken with fixed unrenormalised<br />
parameters)<br />
γ ˜S<br />
= µ ∂<br />
∂µ ∣ ln Z ˜S<br />
, γ S = µ ∂<br />
0<br />
∂µ ∣ ln Z S , (7.66)<br />
0<br />
γ τ = µ ∂<br />
∂µ ∣ ln(τ R /τ) =−2+µ ∂<br />
0<br />
∂µ ∣ ln Z τ , (7.67)<br />
0<br />
γ D = µ ∂<br />
∂µ ∣ ln(D R /D) = 1 ( )<br />
γS − γ<br />
0<br />
2<br />
˜S , (7.68)<br />
where we have used the relation (7.59); for model B, Eq. (7.61) gives <strong>in</strong> addition<br />
γ D = γ S = −γ ˜S<br />
. (7.69)<br />
We have also <strong>in</strong>troduced the RG beta function for the nonl<strong>in</strong>ear coupl<strong>in</strong>g u,<br />
β u = µ<br />
∂<br />
(<br />
∂µ ∣ u R = u R d − 4+µ<br />
∂<br />
)<br />
0<br />
∂µ ∣ ln Z u . (7.70)<br />
0<br />
Explicitly, Eqs. (7.63) and (7.62) yield to lowest nontrivial order, with ɛ =<br />
4 − d,<br />
γ S = − n +2<br />
72 u2 R + O(u 3 R) , (7.71)<br />
a =0 : γ D = n +2 (6ln 4 )<br />
72 3 − 1 u 2 R + O(u 3 R) , (7.72)<br />
γ τ = −2+ n +2 u R + O(u 2<br />
6<br />
R) , (7.73)<br />
[<br />
β u = u R −ɛ + n +8<br />
]<br />
u R + O(u 2<br />
6<br />
R) . (7.74)<br />
In the RG equation for the renormalised dynamic susceptibility, Eq. (7.60)<br />
tells us that the second term <strong>in</strong> Eq. (7.65) is to be replaced with −γ S . Its explicit<br />
dependence on the scale µ can be factored out via χ R (µ, D R ,τ R ,u R , q,ω)<br />
= µ −2 ˆχ R<br />
(<br />
τR ,u R , q/µ, ω/D R µ 2+a) , see Eq. (7.55), whence<br />
[<br />
−2 − γ S + γ D D R<br />
∂<br />
∂D R<br />
+ γ τ τ R<br />
∂<br />
∂τ R<br />
+ β u<br />
]<br />
∂<br />
ˆχ R (D R ,τ R ,u R )=0. (7.75)<br />
∂u R<br />
This l<strong>in</strong>ear partial differential equation is readily solved by means of the<br />
method of characteristics, as is Eq. (7.65) for the vertex functions. The idea
314 U. C. Täuber<br />
is to f<strong>in</strong>d a curve parametrisation µ(l) =µl <strong>in</strong> the space spanned by the<br />
parameters ˜D, ˜τ, andũ such that<br />
l d ˜D(l)<br />
dl<br />
= ˜D(l) γ D (l) , l d˜τ(l)<br />
dl<br />
= ˜τ(l) γ τ (l) , l dũ(l)<br />
dl<br />
= β u (l) , (7.76)<br />
with <strong>in</strong>itial values D R , τ R ,andu R , respectively at l =1.Thefirst-order ord<strong>in</strong>ary<br />
differential equations (7.76), with γ D (l) =γ D (ũ(l)) etc. def<strong>in</strong>e runn<strong>in</strong>g<br />
coupl<strong>in</strong>gs that describe how the parameters of the theory change under scale<br />
transformations µ → µl. The formal solutions for ˜D(l) and˜τ(l) read<br />
˜D(l) =D R exp<br />
[ ∫ l<br />
1<br />
γ D (l ′ ) dl′<br />
l ′ ]<br />
, ˜τ(l) =τ R exp<br />
[ ∫ l<br />
1<br />
γ τ (l ′ ) dl′<br />
l ′ ]<br />
. (7.77)<br />
For the function ˆχ(l) =ˆχ R ( ˜D(l), ˜τ(l), ũ(l)), we then obta<strong>in</strong> another ord<strong>in</strong>ary<br />
differential equation, namely<br />
which is solved by<br />
l dˆχ(l)<br />
dl<br />
ˆχ(l) =ˆχ(1) l 2 exp<br />
=[2+γ S (l)] ˆχ(l) , (7.78)<br />
[ ∫ l<br />
1<br />
γ S (l ′ ) dl′<br />
l ′ ]<br />
. (7.79)<br />
Collect<strong>in</strong>g everyth<strong>in</strong>g, we f<strong>in</strong>ally arrive at<br />
[<br />
χ R (µ, D R ,τ R ,u R , q,ω)=(µl) −2 exp<br />
× ˆχ R<br />
(<br />
−<br />
∫ l<br />
1<br />
γ S (l ′ ) dl′<br />
l ′ ]<br />
˜τ(l), ũ(l), |q|<br />
µl , ω<br />
˜D(l)(µl) 2+a )<br />
. (7.80)<br />
The solution (7.80) of the RG equation (7.75), along with the flow equations<br />
(7.76), (7.77) for the runn<strong>in</strong>g coupl<strong>in</strong>gs tell us how the dynamic susceptibility<br />
depends on the (momentum) scale µlat which we consider the theory.<br />
Similar relations can be obta<strong>in</strong>ed for arbitrary vertex functions by solv<strong>in</strong>g the<br />
associated RG equations (7.65) [13]. The po<strong>in</strong>t here is that the right-hand side<br />
of Eq. (7.80) may be evaluated outside the IR-s<strong>in</strong>gular regime, by fix<strong>in</strong>g one<br />
of its arguments at a f<strong>in</strong>ite value, say |q|/µ l = 1. The function ˆχ R is regular,<br />
and can be calculated by means of perturbation theory. A scale-<strong>in</strong>variant<br />
regime is characterised by the renormalised nonl<strong>in</strong>ear coupl<strong>in</strong>g u R becom<strong>in</strong>g<br />
<strong>in</strong>dependent of the scale µl,orũ(l) → u ∗ = const. For an RG fixed po<strong>in</strong>t to<br />
be <strong>in</strong>frared-stable, we thus require<br />
β u (u ∗ )=0, β ′ u(u ∗ ) > 0 , (7.81)
7 Field-Theory Approaches to Nonequilibrium Dynamics 315<br />
s<strong>in</strong>ce Eq. (7.76) then implies that ũ(l → 0) → u ∗ . Tak<strong>in</strong>g the limit l → 0<br />
thus provides the desired mapp<strong>in</strong>g of physical observables such as (7.80) onto<br />
the critical region. In the vic<strong>in</strong>ity of an IR-stable RG fixed po<strong>in</strong>t, Eq. (7.77)<br />
yields the power laws ˜D(l) ≈ D R l γ∗ D , where γ<br />
∗<br />
D = γ D (l → 0) = γ D (u ∗ ), etc.<br />
Consequently, Eq. (7.80) reduces to<br />
(<br />
χ R (τ R , q,ω) ≈ µ −2 l −2−γ∗ S ˆχR τ R l γ∗ τ ,u ∗ , |q|<br />
)<br />
µl , ω<br />
, (7.82)<br />
D R µ 2+a l 2+a+γ∗ D<br />
and upon match<strong>in</strong>g l = |q|/µ we recover the dynamic scal<strong>in</strong>g law (7.12) with<br />
the critical exponents<br />
η = −γ ∗ S , ν = −1/γ ∗ τ , z =2+a + γ ∗ D . (7.83)<br />
To one-loop order, we obta<strong>in</strong> from the RG beta function (7.74)<br />
u ∗ H =<br />
6 ɛ<br />
n +8 + O(ɛ2 ) . (7.84)<br />
Here we have <strong>in</strong>dicated that our perturbative expansion for small u has effectively<br />
turned <strong>in</strong>to a dimensional expansion <strong>in</strong> ɛ = d c − d. In dimensions<br />
d0. With<br />
Eqs. (7.71) and (7.73), the identifications (7.83) then give us explicit results<br />
for the static scal<strong>in</strong>g exponents, as mere functions of dimension d =4− ɛ and<br />
the number of order parameter components n,<br />
η = n +2<br />
2(n +8) 2 ɛ2 + O(ɛ 3 ) ,<br />
1<br />
ν<br />
=2−<br />
n +2<br />
n +8 ɛ + O(ɛ2 ) . (7.85)<br />
For model A with nonconserved order parameter, the two-loop result (7.72)<br />
yields the <strong>in</strong>dependent dynamic critical exponent<br />
a =0 : z =2+cη , c=6ln 4 − 1+O(ɛ) ; (7.86)<br />
3<br />
for model B with conserved order parameter, <strong>in</strong>stead γD ∗ = γ∗ S = −η, whence<br />
we arrive at the exact scal<strong>in</strong>g relation<br />
a =2 : z =4− η. (7.87)<br />
In dimensions d > d c = 4, the Gaussian fixed po<strong>in</strong>t u ∗ 0 = 0 is stable<br />
(β ′ u(0) = −ɛ >0). Therefore all anomalous dimensions disappear, i.e., γ ∗ S =<br />
0=γ ∗ D and γ∗ τ = −2, and we are left with the mean-field critical exponents<br />
η 0 =0,ν 0 =1/2, and z 0 =2+a. Precisely at the upper critical dimension<br />
d c = 4, the RG flow equation for the nonl<strong>in</strong>ear coupl<strong>in</strong>g becomes<br />
which is solved by<br />
l dũ(l)<br />
dl<br />
= n +8<br />
6<br />
(<br />
ũ(l) 2 + O ũ(l) 3) , (7.88)
316 U. C. Täuber<br />
ũ(l) =<br />
u R<br />
1 − n+8<br />
6<br />
u R ln l . (7.89)<br />
In four dimensions, ũ(l) → 0, but only logarithmically slowly, which causes<br />
logarithmic corrections to the mean-field critical power laws. For example,<br />
upon <strong>in</strong>sert<strong>in</strong>g Eq. (7.89) <strong>in</strong>to the flow equation (7.76), one f<strong>in</strong>ds ˜τ(l) ∼<br />
τ R l −2 (ln |l|) −(n+2)/(n+8) ; with ˜τ(l = ξ −1 )=O(1), iterative <strong>in</strong>version yields<br />
ξ(τ R ) ∼ τ −1/2<br />
R<br />
(ln τ R ) (n+2)/2(n+8) . (7.90)<br />
This concludes our derivation of asymptotic scal<strong>in</strong>g laws for the critical<br />
dynamics of the purely relaxational models A and B, and the explicit computation<br />
of the scal<strong>in</strong>g exponents <strong>in</strong> powers of ɛ = d c − d. In the follow<strong>in</strong>g<br />
sections, I will briefly sketch how the response functional formalism and the<br />
dynamic renormalisation group can be employed to study the critical dynamics<br />
of systems with reversible mode-coupl<strong>in</strong>g terms, the “age<strong>in</strong>g’ behaviour<br />
<strong>in</strong>duced by quench<strong>in</strong>g from random <strong>in</strong>itial conditions to the critical po<strong>in</strong>t, the<br />
effects of violat<strong>in</strong>g the detailed balance constra<strong>in</strong>ts on universal dynamic critical<br />
properties, and the generically scale-<strong>in</strong>variant features of nonequilibrium<br />
systems such as driven diffusive Is<strong>in</strong>g lattice gases.<br />
7.1.6 Critical Dynamics with Reversible Mode-Coupl<strong>in</strong>gs<br />
In the previous chapters, we have assumed purely relaxational dynamics for<br />
the order parameter, see Eq. (7.16). In general, however, there are also reversible<br />
contributions to the systematic force terms F α that enter its Langev<strong>in</strong><br />
equation [7,24]. Consider the Hamiltonian dynamics of microscopic variables,<br />
say, local sp<strong>in</strong> densities, at T =0:∂ t Sm(x, α t) = {H[S m ],Sm(x, α t)}. Here,<br />
the Poisson brackets {A, B} constitute the classical analog of the quantummechanical<br />
commutator<br />
i <br />
[A, B] (correspondence pr<strong>in</strong>ciple). Upon coarsegra<strong>in</strong><strong>in</strong>g,<br />
the microscopic variables Sm α become the mesoscopic hydrodynamic<br />
fields S α . S<strong>in</strong>ce the set of slow modes should provide a complete description<br />
of the critical dynamics, we may formally expand<br />
{<br />
}<br />
H[S],S α (x) =<br />
∫<br />
d d x ∑ ′ β<br />
δH[S]<br />
δS β (x ′ ) Qβα (x ′ , x) , (7.91)<br />
with the mutual Poisson brackets of the hydrodynamic variables<br />
{<br />
}<br />
Q αβ (x, x ′ )= S α (x),S β (x ′ ) = −Q βα (x ′ , x) . (7.92)<br />
By <strong>in</strong>spection of the associated Fokker–Planck equation, one may then<br />
establish an additional equilibrium condition <strong>in</strong> order for the time-dependent<br />
probability distribution to reach the canonical limit (7.6): P[S, t] →P eq [S] as<br />
t →∞provided the probability current is divergence-free <strong>in</strong> the space spanned<br />
by the stochastic fields S α (x):
7 Field-Theory Approaches to Nonequilibrium Dynamics 317<br />
∫<br />
d d x ∑ δ<br />
(<br />
δS α F α<br />
(x)<br />
rev[S] e −H[S]/kBT =0.<br />
α<br />
(7.93)<br />
It turns out that this equilibrium condition is often more crucial than the<br />
E<strong>in</strong>ste<strong>in</strong> relation (7.17). In order to satisfy Eq. (7.93) at T ≠ 0, we must<br />
supplement Eq. (7.91) by a f<strong>in</strong>ite-temperature correction, whereupon the reversible<br />
mode-coupl<strong>in</strong>g contributions to the systematic forces become<br />
∫<br />
Frev[S](x) α =− d d x ∑ ′ β<br />
[<br />
Q αβ (x, x ′ ) δH[S]<br />
δS β (x ′ ) − k BT δQαβ (x, x ′ ]<br />
)<br />
δS β (x ′ ,<br />
)<br />
(7.94)<br />
and the complete coupled set of stochastic differential equations reads<br />
∂S α (x,t)<br />
∂t<br />
= F α rev[S](x,t) − D α (i∇) aα δH[S]<br />
δS α (x,t) + ζα (x,t) , (7.95)<br />
where as before the D α denote the relaxation coefficients, and a α =0or2<br />
respectively for nonconserved and conserved modes.<br />
As an <strong>in</strong>structive example, let us consider the Heisenberg model for<br />
isotropic ferromagnets, H[{S j }]=− 1 ∑ N<br />
2 j,k=1 J jk S j · S k , where the sp<strong>in</strong> operators<br />
satisfy the usual commutation relations [Sj α,Sβ k ]=i ∑ γ ɛαβγ S γ j δ jk.<br />
The correspond<strong>in</strong>g Poisson brackets for the magnetisation density read<br />
Q αβ (x, x ′ )=−g ∑ γ<br />
ɛ αβγ S γ (x) δ(x − x ′ ) , (7.96)<br />
where the purely dynamical coupl<strong>in</strong>g g <strong>in</strong>corporates various factors that<br />
emerge upon coarse-gra<strong>in</strong><strong>in</strong>g and tak<strong>in</strong>g the cont<strong>in</strong>uum limit. The second<br />
contribution <strong>in</strong> Eq. (7.94) vanishes, s<strong>in</strong>ce it reduces to a contraction of the<br />
antisymmetric tensor ɛ αβγ with the Kronecker symbol δ βγ , whence we arrive<br />
at the Langev<strong>in</strong> equations govern<strong>in</strong>g the critical dynamics of the three order<br />
parameter components for isotropic ferromagnets [25]<br />
∂S(x,t)<br />
∂t<br />
= −g S(x,t) × δH[S]<br />
δS(x,t) + D δH[S]<br />
∇2 + ζ(x,t) , (7.97)<br />
δS(x,t)<br />
with 〈ζ(x,t)〉 = 0. S<strong>in</strong>ce [H[{S j }], ∑ k Sα k<br />
] = 0, the total magnetisation is<br />
conserved, whence the noise correlators should be taken as<br />
〈<br />
ζ α (x,t) ζ β (x ′ ,t ′ ) 〉 = −2Dk B T ∇ 2 δ(x − x ′ ) δ(t − t ′ ) δ αβ . (7.98)<br />
The vector product term <strong>in</strong> Eq. (7.97) describes the sp<strong>in</strong> precession <strong>in</strong> the local<br />
effective magnetic field δH[S]/δS, which <strong>in</strong>cludes a contribution <strong>in</strong>duced by<br />
the exchange <strong>in</strong>teraction.<br />
The Langev<strong>in</strong> equation (7.97) and (7.98) with the Hamiltonian (7.7) for<br />
n = 3 def<strong>in</strong>e the so-called model J [7]. In addition to the model B response
318 U. C. Täuber<br />
functional (7.35) and (7.36) with a = 2 (sett<strong>in</strong>g k B T = 1 aga<strong>in</strong>), the reversible<br />
force <strong>in</strong> Eq. (7.97) leads to an additional contribution to the action<br />
∫ ∫<br />
A mc [ ˜S,S]=−g d d x dt ∑ ɛ αβγ ˜Sα S β ( ∇ 2 S γ + h γ) , (7.99)<br />
α,β,γ<br />
which gives rise to an additional mode-coupl<strong>in</strong>g vertex, as depicted <strong>in</strong> Fig. 7.5(a).<br />
Power count<strong>in</strong>g yields the scal<strong>in</strong>g dimension [g] =µ 3−d/2 for the associated<br />
coupl<strong>in</strong>g strength, whence we expect a dynamical upper critical dimension<br />
d ′ c = 6. However, s<strong>in</strong>ce we are <strong>in</strong>vestigat<strong>in</strong>g a system <strong>in</strong> thermal equilibrium,<br />
we can treat its thermodynamics and static properties separately from its dynamics.<br />
Obviously therefore, the static critical exponents must still be given<br />
(to lowest nontrivial order and for d
7 Field-Theory Approaches to Nonequilibrium Dynamics 319<br />
where B d = Γ (4 − d/2)/2 d dπ d/2 , Eq. (7.101) implies the identity [9]<br />
Z g = Z S . (7.103)<br />
For the RG beta function associated with the effective coupl<strong>in</strong>g enter<strong>in</strong>g the<br />
loop corrections, we thus <strong>in</strong>fer<br />
β f = µ ∂<br />
∂µ ∣ f R = f R (d − 6+γ S − 2 γ D ) . (7.104)<br />
0<br />
Consequently, at any nontrivial IR-stable RG fixed po<strong>in</strong>t 0
320 U. C. Täuber<br />
7.1.7 Critical Relaxation, Initial Slip, and Age<strong>in</strong>g<br />
We beg<strong>in</strong> with a brief discussion of the coarsen<strong>in</strong>g dynamics of systems described<br />
by model A/B k<strong>in</strong>etics that are rapidly quenched from a disordered<br />
state at T ≫ T c to the critical po<strong>in</strong>t T ≈ T c [27, 28]. The situation may be<br />
modeled as a relaxation from Gaussian random <strong>in</strong>itial conditions, i.e., the<br />
probability distribution for the order parameter at t = 0 can be taken as<br />
(<br />
P[S, t =0]∝ e −H0[S] =exp − ∆ ∫<br />
d d x ∑ )<br />
[S α (x, 0) − a α (x)] 2 , (7.108)<br />
2<br />
α<br />
where the functions a α (x) specify the most likely <strong>in</strong>itial configurations. Power<br />
count<strong>in</strong>g for the parameter ∆ gives [∆] =µ 2 , whence it is a relevant perturbation<br />
that will flow to ∆ →∞under the RG. Asymptotically, therefore, the<br />
system will be governed by sharp Dirichlet boundary conditions. Whereas the<br />
response propagators rema<strong>in</strong>s a causal function of the time difference between<br />
applied perturbation and effect, G 0 (q,t−t ′ )=Θ(t−t ′ )e −Dqa (r+q 2 )(t−t ′) ,see<br />
Eq. (7.23), time translation <strong>in</strong>variance is broken by the <strong>in</strong>itial state <strong>in</strong> the<br />
Dirichlet correlator of the Gaussian model,<br />
C D (q; t, t ′ )= 1 (<br />
r + q 2 e −Dqa (r+q 2 ) |t−t ′| ) − e −Dqa (r+q 2 )(t+t ′ )<br />
. (7.109)<br />
Away from criticality, i.e., for r>0andq ≠ 0, temporal correlations decay<br />
exponentially fast, and the system quickly approaches the stationary equilibrium<br />
state. However, as T → T c , the equilibration time diverges accord<strong>in</strong>g to<br />
t c ∼|τ| −zν →∞, and the system never reaches thermal equilibrium. Twotime<br />
correlation functions will then depend on both times separately, <strong>in</strong> a<br />
specific manner to be addressed below, a phenomenon termed critical “age<strong>in</strong>g”<br />
(for more details, see Refs. [29, 30]).<br />
The field-theoretic treatment of the model A/B dynamical action (7.35),<br />
(7.36) with the <strong>in</strong>itial term (7.108) follows the theory of boundary critical<br />
phenomena [31]. However, it turns out that additional s<strong>in</strong>gularities on the<br />
temporal “surface” at t + t ′ = 0 appear only for model A, and can be <strong>in</strong>corporated<br />
<strong>in</strong>to a s<strong>in</strong>gle new renormalisation factor; to one-loop order, one<br />
f<strong>in</strong>ds [27, 28]<br />
a =0 : ˜Sα R (x, 0) = (Z 0 Z ˜S) 1/2 ˜Sα (x, 0) , Z 0 =1− n +2<br />
6<br />
u R<br />
ɛ<br />
. (7.110)<br />
This <strong>in</strong> turn leads to a s<strong>in</strong>gle <strong>in</strong>dependent critical exponent associated with<br />
the <strong>in</strong>itial time relaxation, the <strong>in</strong>itial slip exponent, which becomes for the<br />
purely relaxational models A and B with nonconserved and conserved order<br />
parameter:<br />
a =0 : θ = γ∗ 0<br />
2 z = n +2<br />
4(n +8) ɛ + O(ɛ2 ) , a =2 : θ =0. (7.111)
7 Field-Theory Approaches to Nonequilibrium Dynamics 321<br />
In order to obta<strong>in</strong> the short-time scal<strong>in</strong>g laws for the dynamic response and<br />
correlation functions <strong>in</strong> the age<strong>in</strong>g limit t ′ /t → 0, one requires additional<br />
<strong>in</strong>formation that can be garnered from the short-distance operator product<br />
expansion for the fields,<br />
t → 0: ˜S(x,t)=˜σ(t) ˜S0 (x) , S(x,t)=σ(t) ˜S 0 (x) . (7.112)<br />
Subsequent analysis then yields eventually [27, 28]<br />
χ(q; t, t ′ → 0) = |q| z−2+η ( t<br />
t ′ ) θ<br />
ˆχ 0 (q ξ,|q| z Dt) , (7.113)<br />
C(q; t, t ′ → 0) = |q| −2+η ( t<br />
t ′ ) θ−1<br />
Ĉ0 (q ξ,|q| z Dt) , (7.114)<br />
and for the time dependence of the mean order parameter<br />
( )<br />
〈S(t)〉 = S 0 t θ′ Ŝ S 0 t θ′ +β/zν<br />
, (7.115)<br />
a =0 : θ ′ = θ − z − 2+η<br />
z<br />
, a =2 : θ ′ = θ =0. (7.116)<br />
One may also compute the universal fluctuation–dissipation ratios <strong>in</strong> this<br />
nonequilibrium age<strong>in</strong>g regime [29, 30]. It emerges, though, that these depend<br />
on the quantity under <strong>in</strong>vestigation, which prohibits a unique def<strong>in</strong>ition of an<br />
effective nonequilibrium temperature for critical age<strong>in</strong>g. The method sketched<br />
above can be extended to models with reversible mode-coupl<strong>in</strong>gs [32]. For<br />
model J captur<strong>in</strong>g the critical dynamics of isotropic ferromagnets, one f<strong>in</strong>ds<br />
θ = z − 4+η<br />
z<br />
= − 6 − d − η<br />
d +2− η ; (7.117)<br />
<strong>in</strong> systems where a nonconserved order parameter is dynamically coupled to<br />
other conserved modes, the <strong>in</strong>itial slip exponent θ is actually not a universal<br />
number, but depends on the width of the <strong>in</strong>itial distribution [32].<br />
7.1.8 Nonequilibrium Relaxational Critical Dynamics<br />
Next we address the question [33], What happens if the detailed balance conditions<br />
(7.17) and (7.93) are violated? To start, we change the noise strength<br />
D → ˜D <strong>in</strong> the purely relaxational models A and B, which (<strong>in</strong> our units) violates<br />
the E<strong>in</strong>ste<strong>in</strong> relation (7.17). However, this modification can obviously be<br />
absorbed <strong>in</strong>to a rescaled effective temperature, k B T → k B T ′ = ˜D/D. Formally<br />
this is established by means of the dynamical action (7.34), which now reads<br />
∫ ∫<br />
A[ ˜S,S] = d d x dt ∑ ˜S<br />
[∂ α t S α + D (i∇) a ( r − ∇ 2) S α<br />
α<br />
− ˜D (i∇) a ˜Sα + D u 6 (i∇)a S ∑ ]<br />
α S β S β . (7.118)<br />
β
322 U. C. Täuber<br />
√<br />
Upon simple rescal<strong>in</strong>g ˜S α → ˜S ′α = ˜S α ˜D/D, S α → S ′α = S<br />
√D/ α ˜D, the<br />
response functional (7.118) recovers its equilibrium form, albeit with modified<br />
nonl<strong>in</strong>ear coupl<strong>in</strong>g u → ũ = u ˜D/D. However, the universal asymptotic<br />
properties of these models are governed by the Heisenberg fixed po<strong>in</strong>t (7.84),<br />
and the specific value of the (renormalised) coupl<strong>in</strong>g, which only serves as<br />
the <strong>in</strong>itial condition for the RG flow, does not matter. In fact, the relaxational<br />
dynamics of the k<strong>in</strong>etic Is<strong>in</strong>g model with Glauber dynamics (model<br />
A with n = 1) is known to be quite stable aga<strong>in</strong>st nonequilibrium perturbations<br />
[34, 35], even if these break the Is<strong>in</strong>g Z 2 symmetry [36]. For model J<br />
the above rescal<strong>in</strong>g modifies <strong>in</strong> a similar manner √ merely the mode-coupl<strong>in</strong>g<br />
strength <strong>in</strong> Eq. (7.99), namely g → ˜g = g ˜D/D [37]. Aga<strong>in</strong>, s<strong>in</strong>ce the dynamic<br />
critical behaviour is governed by the universal fixed po<strong>in</strong>t (7.107), thermal<br />
equilibrium becomes effectively restored at criticality. More generally, it<br />
has been established that isotropic detailed balance violations do not affect<br />
the universal properties <strong>in</strong> other models for critical dynamics that conta<strong>in</strong><br />
additional conserved variables either: the equilibrium RG fixed po<strong>in</strong>ts tend to<br />
be asymptotically stable [33].<br />
In systems with conserved order parameter, however, we may <strong>in</strong> addition<br />
<strong>in</strong>troduce spatially anisotropic violations of E<strong>in</strong>ste<strong>in</strong>’s relation; for example, <strong>in</strong><br />
model B one can allow for anisotropic relaxation −D ∇ 2 →−D ⊥ ∇ 2 ⊥ −D ‖ ∇ 2 ‖ ,<br />
with different rates <strong>in</strong> two spatial subsectors and concomitantly anisotropic<br />
noise correlations − ˜D ∇ 2 →−˜D ⊥ ∇ 2 ⊥ − ˜D ‖ ∇ 2 ‖. We have thus produced a<br />
truly nonequilibrium situation provided ˜D ⊥ /D ⊥ ≠ ˜D ‖ /D ‖ , which we may<br />
<strong>in</strong>terpret as hav<strong>in</strong>g effectively coupled the longitud<strong>in</strong>al and transverse spatial<br />
sectors to heat baths with different temperatures T ⊥
7 Field-Theory Approaches to Nonequilibrium Dynamics 323<br />
Quite remarkably, the Langev<strong>in</strong> equation (7.119) can be derived as an equilibrium<br />
diffusive relaxational k<strong>in</strong>etics<br />
∂S α (x,t)<br />
∂t<br />
= D ∇ 2 ⊥<br />
from an effective long-range Hamiltonian<br />
∫<br />
H eff [S] =<br />
+ ũ<br />
4!<br />
δH eff [S]<br />
δS α (x,t) + ζα (x,t) (7.121)<br />
d d q q 2 ⊥ (r + q2 ⊥ )+c q2 ∑<br />
‖<br />
∣ S α<br />
(2π) d 2 q 2 (q) ∣ 2<br />
⊥<br />
α<br />
∫<br />
d d x ∑ [S α (x)] 2 [S β (x)] 2 . (7.122)<br />
α,β<br />
Power count<strong>in</strong>g gives [ũ] =µ 4−d ‖−d : the spatial anisotropy suppresses longitud<strong>in</strong>al<br />
fluctuations and lower the upper critical dimension to d c =4−d ‖ .The<br />
anisotropic correlations encoded <strong>in</strong> Eq. (7.122) also reduce the lower critical<br />
dimension and affect the nature of the ordered phase [39, 41].<br />
The scal<strong>in</strong>g law for, e.g., the dynamic response function takes the form<br />
( √ )<br />
τ c<br />
χ(τ ⊥ , q ⊥ , q ‖ ,ω)=|q ⊥ | −2+η ˆχ<br />
|q ⊥ | , |q‖ |<br />
1/ν |q ⊥ | 1+∆ , ω<br />
D |q ⊥ | z , (7.123)<br />
where we have <strong>in</strong>troduced a new anisotropy exponent ∆. S<strong>in</strong>ce the nonl<strong>in</strong>ear<br />
coupl<strong>in</strong>g ũ only affects the transverse sector, we f<strong>in</strong>d to all orders <strong>in</strong> the<br />
perturbation expansion:<br />
Γ (1,1) (q ⊥ =0, q ‖ ,ω)=iω + Dcq 2 ‖ , (7.124)<br />
and consequently obta<strong>in</strong> the Z factor identity<br />
Z c = Z −1<br />
D<br />
= Z−1 S<br />
, (7.125)<br />
which at any IR-stable RG fixed po<strong>in</strong>t implies the exact scal<strong>in</strong>g relations<br />
z =4− η,<br />
∆=1− γ∗ c<br />
2 =1− η 2 = z 2 − 1 , (7.126)<br />
whereas the scal<strong>in</strong>g exponents for the longitud<strong>in</strong>al sector read<br />
z ‖ =<br />
z<br />
1+∆ =2, ν ‖ = ν (1 + ∆) = ν (4 − η) . (7.127)<br />
2<br />
As for the equilibrium model B, the only <strong>in</strong>dependent critical exponents to<br />
be determ<strong>in</strong>ed are η and ν. To one-loop order, only the comb<strong>in</strong>atorics of the<br />
Feynman diagrams (see Fig. 7.2) enters their explicit values, whence one f<strong>in</strong>ds<br />
for d
324 U. C. Täuber<br />
η =0+O(ɛ 2 ) ,<br />
1<br />
ν<br />
=2−<br />
n +2<br />
n +8 ɛ + O(ɛ2 ) , (7.128)<br />
albeit with different ɛ =4−d−d ‖ . To two-loop order, however, the anisotropy<br />
manifestly affects the evaluation of the loop contributions, and the value for<br />
η deviates from the expression <strong>in</strong> Eq. (7.85) [40].<br />
Interest<strong>in</strong>gly, an analogously constructed nonequilibrium two-temperature<br />
model J with reversible mode-coupl<strong>in</strong>g vertex cannot be cast <strong>in</strong>to a form that<br />
is equivalent to an equilibrium system, for ow<strong>in</strong>g to the emerg<strong>in</strong>g anisotropy,<br />
the condition (7.93) cannot be satisfied. A one-loop RG analysis yields a runaway<br />
flow, and no stable RG fixed po<strong>in</strong>t is found [38]. Similar behaviour ensues<br />
<strong>in</strong> other anisotropic nonequilibrium variants of critical dynamics models with<br />
conserved order parameter; the precise <strong>in</strong>terpretation of the apparent <strong>in</strong>stability<br />
is as yet unclear [33].<br />
7.1.9 Driven Diffusive Systems<br />
F<strong>in</strong>ally, we wish to consider Langev<strong>in</strong> representations of genu<strong>in</strong>ely nonequilibrium<br />
systems, namely driven diffusive lattice gases (for a comprehensive<br />
overview, see Ref. [42]). First we address the coarse-gra<strong>in</strong>ed cont<strong>in</strong>uum version<br />
of the asymmetric exclusion process, i.e., hard-core repulsive particles that hop<br />
preferentially <strong>in</strong> one direction. We describe this system <strong>in</strong> terms of a conserved<br />
particle density, whose fluctuations we denote with S(x,t), such that 〈S〉 =0,<br />
obey<strong>in</strong>g a cont<strong>in</strong>uity equation ∂ t S(x,t)+∇ · J(x,t) = 0. We assume the system<br />
to be driven along the “‖’ direction; <strong>in</strong> the transverse sector (of dimension<br />
d ⊥ = d − 1) we thus just have a noisy diffusion current J ⊥ = −D ∇ ⊥ S + η,<br />
whereas there is a nonl<strong>in</strong>ear term, stemm<strong>in</strong>g from the hard-core <strong>in</strong>teractions,<br />
<strong>in</strong> the current along the direction of the external drive, with J 0 ‖ =const.: J ‖ =<br />
J 0 ‖ −Dc∇ ‖ S − 1 2 Dg S2 +η ‖ . For the stochastic currents, we assume Gaussian<br />
white noise 〈η i 〉 =0=〈η ‖ 〉 and 〈η i (x,t) η j (x ′ ,t ′ )〉 =2Dδ(x − x ′ ) δ(t − t ′ ) δ ij ,<br />
〈<br />
η‖ (x,t) η ‖ (x ′ ,t ′ ) 〉 =2D ˜cδ(x − x ′ ) δ(t − t ′ ). Notice that s<strong>in</strong>ce we are not <strong>in</strong><br />
thermal equilibrium, E<strong>in</strong>ste<strong>in</strong>’s relation need not be fulfilled. We can however<br />
always rescale the field to satisfy it <strong>in</strong> the transverse sector; the ratio w =˜c/c<br />
then measures the deviation from equilibrium. These considerations yield the<br />
generic Langev<strong>in</strong> equation for the density fluctuations <strong>in</strong> driven diffusive systems<br />
(DDS) [43, 44]<br />
∂S(x,t)<br />
∂t<br />
( )<br />
= D ∇ 2 ⊥ + c ∇ 2 ‖ S(x,t)+ Dg<br />
2 ∇ ‖S(x,t) 2 + ζ(x,t) , (7.129)<br />
with conserved noise ζ = −∇ ⊥ · η −∇ ‖ η ‖ , where 〈ζ〉 =0and<br />
( )<br />
〈ζ(x,t) ζ(x ′ ,t ′ )〉 = −2D ∇ 2 ⊥ +˜c ∇ 2 ‖ δ(x − x ′ ) δ(t − t ′ ) . (7.130)<br />
Notice that the drive term ∝ g breaks both the system’s spatial reflection<br />
symmetry and the Is<strong>in</strong>g Z 2 symmetry S →−S.
7 Field-Theory Approaches to Nonequilibrium Dynamics 325<br />
The correspond<strong>in</strong>g Janssen–De Dom<strong>in</strong>icis response functional (7.34) reads<br />
∫ ∫ [<br />
A[ ˜S,S] = d d x dt ˜S ∂S<br />
( )<br />
∂t − D ∇ 2 ⊥ + c ∇ 2 ‖ S<br />
]<br />
( )<br />
+ D ∇ 2 ⊥ +˜c ∇ 2 Dg<br />
‖ ˜S −<br />
2 ∇ ‖ S 2 . (7.131)<br />
It describes a “massless” theory, hence we expect the system to be generically<br />
scale-<strong>in</strong>variant, without the need to tune it to a special po<strong>in</strong>t <strong>in</strong> parameter<br />
space. The nonl<strong>in</strong>ear drive term will <strong>in</strong>duce anomalous scal<strong>in</strong>g <strong>in</strong> the drive<br />
direction, different from ord<strong>in</strong>ary diffusive behaviour. In the transverse sector,<br />
however, we have to all orders <strong>in</strong> the perturbation expansion simply<br />
Γ (1,1) (q ⊥ ,q ‖ =0,ω)=iω +D q 2 ⊥ , Γ (2,0) (q ⊥ ,q ‖ =0,ω)=−2D q 2 ⊥ , (7.132)<br />
s<strong>in</strong>ce the nonl<strong>in</strong>ear three-po<strong>in</strong>t vertex, which is of the form depicted <strong>in</strong><br />
Fig. 7.5(a), is proportional to iq ‖ . Consequently,<br />
which immediately implies<br />
Z ˜S<br />
= Z S = Z D =1, (7.133)<br />
η =0, z =2. (7.134)<br />
Moreover, the nonl<strong>in</strong>ear coupl<strong>in</strong>g g itself does not renormalise either as a<br />
consequence of Galilean <strong>in</strong>variance. Namely, the Langev<strong>in</strong> equation (7.129)<br />
and the action (7.131) are left <strong>in</strong>variant under Galilean transformations<br />
S ′ (x ′ ⊥,x ′ ‖ ,t′ )=S(x ⊥ ,x ‖ − Dgv t, t) − v ; (7.135)<br />
thus, the boost velocity v must scale as the field S under renormalisation, and<br />
s<strong>in</strong>ce the product Dgv must be <strong>in</strong>variant under the RG, this leaves us with<br />
Z g = Z −1<br />
D<br />
Z−1 S<br />
=1. (7.136)<br />
The effective nonl<strong>in</strong>ear coupl<strong>in</strong>g govern<strong>in</strong>g the perturbation expansion <strong>in</strong><br />
terms of loop diagrams turns out to be g 2 /c 3/2 ; if we def<strong>in</strong>e its renormalised<br />
counterpart as<br />
v R = Zc 3/2 vC d µ d−2 , (7.137)<br />
with the convenient choice C d = Γ (2 − d/2)/2 d−1 π d/2 , we see that the associated<br />
RG beta function becomes<br />
β v = v R<br />
(d − 2 − 3 )<br />
2 γ c . (7.138)<br />
At any nontrivial RG fixed po<strong>in</strong>t 0 < v ∗ < ∞, therefore γ ∗ c = 2 3<br />
(d − 2).<br />
We thus <strong>in</strong>fer that below the upper critical dimension d c = 2 for DDS, the<br />
longitud<strong>in</strong>al scal<strong>in</strong>g exponents are fixed by the system’s symmetry [43, 44],
326 U. C. Täuber<br />
∆ = − γ∗ c<br />
2 = 2 − d<br />
3<br />
, z ‖ = 2<br />
1+∆ = 6<br />
5 − d . (7.139)<br />
An explicit one-loop calculation for the two-po<strong>in</strong>t vertex functions, see<br />
Fig. 7.5(b) and (c), yields<br />
γ c = − v R<br />
16 (3 + w R) , γ˜c = − v R<br />
(<br />
3w<br />
−1<br />
R<br />
32<br />
R)<br />
, (7.140)<br />
β w = w R (γ˜c − γ c )=− v R<br />
32 (w R − 1) (w R − 3) . (7.141)<br />
This establishes that <strong>in</strong> fact the fixed po<strong>in</strong>t w ∗ = 1 is IR-stable (provided 0 <<br />
v ∗ < ∞), which means that asymptotically the E<strong>in</strong>ste<strong>in</strong> relation is satisfied<br />
<strong>in</strong> the longitud<strong>in</strong>al sector as well [43].<br />
In this context, it is <strong>in</strong>structive to make an <strong>in</strong>trigu<strong>in</strong>g connection with the<br />
noisy Burgers equation [45], describ<strong>in</strong>g simplified fluid dynamics <strong>in</strong> terms of<br />
a velocity field u(x,t):<br />
∂u(x,t)<br />
+ Dg<br />
∂t 2 ∇ [ u(x,t) 2] = D ∇ 2 u(x,t)+ζ(x,t) , (7.142)<br />
〈ζ i 〉 =0, 〈ζ i (x,t) ζ j (x ′ ,t ′ )〉 = −2D ∇ i ∇ j δ(x − x ′ ) δ(t − t ′ ) . (7.143)<br />
For Dg = 1, the nonl<strong>in</strong>earity is just the usual fluid advection term. In one<br />
dimension, the Burgers equation (7.142) becomes identical with the DDS<br />
Langev<strong>in</strong> equation (7.129), so we immediately <strong>in</strong>fer its anomalous dynamic<br />
critical exponent z ‖ =3/2. At least <strong>in</strong> one dimension therefore, it should represent<br />
an equilibrium system which asymptotically approaches the canonical<br />
distribution (7.6), where the Hamiltonian is simply the fluid’s k<strong>in</strong>etic energy<br />
(andwehavesetk B T = 1). So let us check the equilibrium condition (7.93)<br />
with P eq [u] ∝ exp [ ∫<br />
− 1 2 u(x) 2 d d x ] :<br />
∫<br />
d d δ<br />
x<br />
δu(x,t) · [∇u(x,t) 2] e − 1 ∫<br />
2 u(x ′ ,t) 2 d d x ′<br />
=<br />
∫ [2∇<br />
· u(x,t) − u(x,t) · ∇u(x,t)<br />
2 ] d d x e − 1 2<br />
∫<br />
u(x ′ ,t) 2 d d x ′ .<br />
With appropriate boundary conditions, the first term here vanishes, but the<br />
second one does so only <strong>in</strong> d = 1: − ∫ u (du 2 /dx)dx = ∫ ∫ u 2 (du/dx)dx =<br />
1<br />
3 (du 3 /dx)dx = 0. Driven diffusive systems <strong>in</strong> one dimension are therefore<br />
subject to a “hidden” fluctuation–dissipation theorem.<br />
To conclude this part on Langev<strong>in</strong> dynamics, let us briefly consider the<br />
driven model B or critical DDS [42], which corresponds to a driven Is<strong>in</strong>g lattice<br />
gas near its critical po<strong>in</strong>t. Here, a conserved scalar field S undergoes a secondorder<br />
phase transition, but similar to the randomly driven case, aga<strong>in</strong> only the<br />
transverse sector is critical. Upon add<strong>in</strong>g the DDS drive term from Eq. (7.129)<br />
to the Langev<strong>in</strong> equation (7.119), we obta<strong>in</strong>
∂S(x,t)<br />
∂t<br />
7 Field-Theory Approaches to Nonequilibrium Dynamics 327<br />
[<br />
= D ∇ 2 ( ) ]<br />
⊥ r − ∇<br />
2<br />
⊥ + c ∇<br />
2<br />
‖ S(x,t)+ D ũ<br />
6 ∇2 ⊥ S(x,t) 3<br />
+ Dg<br />
2 ∇ ‖ S(x,t) 2 + ζ(x,t) , (7.144)<br />
with the (scalar) noise specified <strong>in</strong> Eq. (7.120). The response functional thus<br />
becomes<br />
∫ ∫ [<br />
A[ ˜S,S] = d d x dt ˜S ∂S<br />
[<br />
∂t − D ∇ 2 ( ) ]<br />
⊥ r − ∇<br />
2<br />
⊥ + c ∇<br />
2<br />
‖ S<br />
(<br />
+D ∇ 2 ˜S ⊥ − ũ<br />
6 ∇2 ⊥ S 3 − g ) ]<br />
2 ∇ ‖ S 2 . (7.145)<br />
Power count<strong>in</strong>g gives [g 2 ] = µ 5−d , so the upper critical dimension here is<br />
d c = 5, and [ũ] =µ 3−d . The nonl<strong>in</strong>earity ∝ ũ is thus irrelevant and can<br />
be omitted if we wish to determ<strong>in</strong>e the asymptotic universal scal<strong>in</strong>g laws;<br />
but recall that it is responsible for the phase transition <strong>in</strong> the system. The<br />
rema<strong>in</strong><strong>in</strong>g vertex is then proportional to iq ‖ , whence Eqs. (7.132) and (7.133)<br />
hold for critical DDS as well, and the transverse critical exponents are just<br />
those of the Gaussian model B,<br />
η =0, ν =1/2 , z =4. (7.146)<br />
In addition, Galilean <strong>in</strong>variance with respect to Eq. (7.135) and therefore<br />
Eq. (7.136) hold as before. With the renormalised nonl<strong>in</strong>ear drive strength<br />
def<strong>in</strong>ed similarly to Eq. (7.137), but a different geometric constant and the<br />
scale factor µ d−5 , the associated RG beta function reads<br />
β v = v R<br />
(d − 5 − 3 )<br />
2 γ c , (7.147)<br />
which aga<strong>in</strong> allows us to determ<strong>in</strong>e the longitud<strong>in</strong>al scal<strong>in</strong>g exponents to all<br />
orders <strong>in</strong> perturbation theory, for d
328 U. C. Täuber<br />
7.2 Reaction–Diffusion Systems<br />
We now turn our attention to stochastic <strong>in</strong>teract<strong>in</strong>g particle systems, whose<br />
microscopic dynamics is def<strong>in</strong>ed through a (classical) master equation. Below,<br />
we shall see how the latter can be mapped onto a stochastic quasi-Hamiltonian<br />
<strong>in</strong> a second-quantised bosonic operator representation [10–12]. Tak<strong>in</strong>g the<br />
cont<strong>in</strong>uum limit on the basis of coherent-state path <strong>in</strong>tegrals then yields a field<br />
theory action that may be analysed by the very same RG methods as described<br />
before <strong>in</strong> Subsects. 7.1.3–7.1.5 (for more details, see the recent overview [12]).<br />
7.2.1 Chemical Reactions and Population Dynamics<br />
Our goal is to study systems of “particles” A,B,... that propagate through<br />
hopp<strong>in</strong>g to nearest neighbors on a d-dimensional lattice, or via diffusion <strong>in</strong><br />
the cont<strong>in</strong>uum. Upon encounter, or spontaneously, with given stochastic rates,<br />
these particles may undergo species changes, annihilate, or produce offspr<strong>in</strong>g.<br />
At large densities, the characteristic time scales of the k<strong>in</strong>etics will be governed<br />
by the reaction rates, and the system is said to be reaction-limited. In contrast,<br />
at low densities, any reactions that require at least two particles to be <strong>in</strong><br />
proximity will be diffusion-limited: the basic time scale will be set by the<br />
hopp<strong>in</strong>g rate or diffusion coefficient.<br />
As a first approximation to the dynamics of such “chemical” reactions,<br />
let us assume homogeneous mix<strong>in</strong>g of each species. We may then hope to be<br />
able to capture the k<strong>in</strong>etics <strong>in</strong> terms of rate equations for each particle concentration<br />
or mean density. Note that such a description neglects any spatial<br />
fluctuations and correlations <strong>in</strong> the system, and is therefore <strong>in</strong> character a<br />
mean-field approximation. As a first illustration consider the annihilation of<br />
k−l >0 particles of species A <strong>in</strong> the irreversible kth-order reaction kA→ lA,<br />
with rate λ. The correspond<strong>in</strong>g rate equation employs a factorisation of the<br />
probability of encounter<strong>in</strong>g k particles at the same po<strong>in</strong>t to simply the kth<br />
power of the concentration a(t),<br />
ȧ(t) =−(k − l) λa(t) k . (7.149)<br />
This ord<strong>in</strong>ary differential equation is readily solved, with the result<br />
k =1 : a(t) =a(0) e −λt , (7.150)<br />
k ≥ 2: a(t) = [ a(0) 1−k +(k − l)(k − 1) λt ] −1/(k−1)<br />
. (7.151)<br />
For simple “radioactive” decay (k = 1), we of course obta<strong>in</strong> an exponential<br />
time dependence, as appropriate for statistically <strong>in</strong>dependent events. For<br />
pair (k = 2) and higher-order (k ≥ 3) processes, however, we f<strong>in</strong>d algebraic<br />
long-time behaviour, a(t) → (λt) −1/(k−1) , with an amplitude that becomes<br />
<strong>in</strong>dependent of the <strong>in</strong>itial density a(0). The absence of a characteristic time<br />
scale h<strong>in</strong>ts at cooperative effects, and we have to ask if and under which circumstances<br />
correlations might qualitatively affect the asymptotic long-time
7 Field-Theory Approaches to Nonequilibrium Dynamics 329<br />
power laws. For accord<strong>in</strong>g to Smoluchowski theory [12], we would expect the<br />
annihilation reactions to produce depletion zones <strong>in</strong> sufficiently low dimensions<br />
d ≤ d c , which would <strong>in</strong> turn <strong>in</strong>duce a considerable slow<strong>in</strong>g down of the<br />
density decay, see Subsect. 7.2.3. For two-species pair annihilation A + B →∅<br />
(without mix<strong>in</strong>g), another complication emerges, namely particle species segregation<br />
<strong>in</strong> dimensions for d ≤ d s ; the regions dom<strong>in</strong>ated by either species<br />
become largely <strong>in</strong>ert, and the annihilation reactions are conf<strong>in</strong>ed to rather<br />
sharp fronts [12].<br />
Competition between particle decay and production processes, e.g., <strong>in</strong> the<br />
reactions A →∅(with rate κ), A ⇋ A + A (with forward and back rates σ<br />
and λ, respectively), leads to even richer scenarios, as can already be <strong>in</strong>ferred<br />
from the associated rate equation<br />
ȧ(t) =(σ − κ) a(t) − λa(t) 2 . (7.152)<br />
For σκ, we encounter an<br />
active state with a(t) → a ∞ =(σ − κ)/λ exponentially, with rate ∼ σ − κ. We<br />
have thus identified a nonequilibrium cont<strong>in</strong>uous phase transition at σ c = κ.<br />
Indeed, as <strong>in</strong> equilibrium critical phenomena, the critical po<strong>in</strong>t is governed by<br />
characteristic power laws; for example, the asymptotic particle density a ∞ ∼<br />
(σ−σ c ) β , and the critical density decay a(t) ∼ (λt) −α with β 0 =1=α 0 <strong>in</strong> the<br />
mean-field approximation. The follow<strong>in</strong>g natural questions then arise: What<br />
are the critical exponents once statistical fluctuations are properly <strong>in</strong>cluded<br />
<strong>in</strong> the analysis? Can we, as <strong>in</strong> equilibrium systems, identify and characterise<br />
certa<strong>in</strong> universality classes, and which microscopic or overall, global features<br />
determ<strong>in</strong>e them and their critical dimension?<br />
Already the previous set of reactions may also be viewed as a (crude) model<br />
for the population dynamics of a s<strong>in</strong>gle species. In the same language, we may<br />
also formulate a stochastic version of the classic Lotka–Volterra predator–prey<br />
competition model [46]: if by themselves, the “predators” A die out accord<strong>in</strong>g<br />
to A →∅, with rate κ, whereas the prey reproduce B → B + B with rate σ,<br />
and thus proliferate with a Malthusian population explosion. The predators<br />
are kept alive and the prey under control through predation, here modeled as<br />
the reaction A + B → A + A: with rate λ, a prey is “eaten” by a predator,<br />
who simultaneously produces an offspr<strong>in</strong>g. The coupled k<strong>in</strong>etic rate equations<br />
for this system read<br />
ȧ(t) =λa(t) b(t) − κa(t) , ḃ(t) =σb(t) − λa(t) b(t) . (7.153)<br />
It is straightforward to show that the quantity K(t) =λ[a(t)+b(t)]−σ ln a(t)−<br />
κ ln b(t) is a constant of motion for this coupled system of differential equations,<br />
i.e., ˙K(t) = 0. As a consequence, the system is governed by regular<br />
population oscillations, whose frequency and amplitude are fully determ<strong>in</strong>ed
330 U. C. Täuber<br />
by the <strong>in</strong>itial conditions. Clearly, this is not a very realistic feature (albeit<br />
mathematically appeal<strong>in</strong>g), and moreover Eqs. (7.153) are known to be quite<br />
unstable with respect to model modifications [46]. Indeed, if one <strong>in</strong>cludes<br />
spatial degrees of freedom and takes account of the full stochasticity of the<br />
processes <strong>in</strong>volved, the system’s behaviour turns out to be much richer [47]:<br />
In the species coexistence phase, one encounters for sufficiently large values of<br />
the predation rate an <strong>in</strong>cessant sequence of “pursuit and evasion” waves that<br />
form quite complex dynamical patterns. In f<strong>in</strong>ite systems, these <strong>in</strong>duce erratic<br />
population oscillations whose features are however <strong>in</strong>dependent of the <strong>in</strong>itial<br />
configuration, but whose amplitude vanishes <strong>in</strong> the thermodynamic limit.<br />
Moreover, if locally the prey “carry<strong>in</strong>g capacity” is limited (correspond<strong>in</strong>g to<br />
restrict<strong>in</strong>g the maximum site occupation number per site on a lattice), there<br />
appears an ext<strong>in</strong>ction threshold for the predator population that separates the<br />
absorb<strong>in</strong>g state of a system filled with prey from the active coexistence regime<br />
through a cont<strong>in</strong>uous phase transition [47].<br />
These examples all call for a systematic approach to <strong>in</strong>clude stochastic<br />
fluctuations <strong>in</strong> the mathematical description of <strong>in</strong>teract<strong>in</strong>g reaction–diffusion<br />
systems that would be conducive to the application of field-theoretic tools, and<br />
thus allow us to br<strong>in</strong>g the powerful mach<strong>in</strong>ery of the dynamic renormalisation<br />
group to bear on these problems. In the follow<strong>in</strong>g, we shall describe such a<br />
general method [48–50] which allows a representation of the classical master<br />
equation <strong>in</strong> terms of a coherent-state path <strong>in</strong>tegral and its subsequent analysis<br />
by means of the RG (for overviews, see Refs. [10–12]).<br />
7.2.2 Field Theory Representation of Master Equations<br />
The above <strong>in</strong>teract<strong>in</strong>g particle systems, when def<strong>in</strong>ed on a d-dimensional lattice<br />
with sites i, are fully characterised by the set of occupation <strong>in</strong>teger numbers<br />
n i =0, 1, 2,... for each particle species. The master equation then describes<br />
the temporal evolution of the configurational probability distribution<br />
P ({n i }; t) through a balance of ga<strong>in</strong> and loss terms. For example, for the b<strong>in</strong>ary<br />
annihilation and coagulation reactions A + A →∅with rate λ and A + A → A<br />
with rate λ ′ , the master equation on a specific site i reads<br />
∂P(n i ; t)<br />
∂t<br />
= λ (n i +2)(n i +1)P (...,n i +2,...; t)<br />
+λ ′ (n i +1)n i P (...,n i +1,...; t)<br />
−(λ + λ ′ ) n i (n i − 1) P (...,n i ,...; t) , (7.154)<br />
with <strong>in</strong>itially P ({n i }, 0) = ∏ i P (n i), e.g., a Poisson distribution P (n i ) =<br />
¯n ni<br />
0 e−¯n0 /n i !. S<strong>in</strong>ce the reactions all change the site occupation numbers by<br />
<strong>in</strong>teger values, a second-quantised Fock space representation is particularly<br />
useful [48–50]. To this end, we <strong>in</strong>troduce the bosonic operator algebra<br />
] [ ] [ ]<br />
[a i ,a j =0= a † i ,a† j , a i ,a † j = δ ij . (7.155)
7 Field-Theory Approaches to Nonequilibrium Dynamics 331<br />
From these commutation relations one establishes <strong>in</strong> the standard manner<br />
that a i and a † i constitute lower<strong>in</strong>g and rais<strong>in</strong>g ladder operators, from which<br />
we may construct the particle number eigenstates |n i 〉,<br />
a i |n i 〉 = n i |n i − 1〉 , a † i |n i〉 = |n i +1〉 , a † i a i |n i 〉 = n i |n i 〉 . (7.156)<br />
(Notice that we have chosen a different normalisation than <strong>in</strong> ord<strong>in</strong>ary quantum<br />
mechanics.) A state with n i particles on sites i is then obta<strong>in</strong>ed from the<br />
empty vaccum state |0〉, def<strong>in</strong>ed through a i |0〉 = 0, as the product state<br />
|{n i }〉 = ∏ ( )<br />
a † ni<br />
i |0〉 . (7.157)<br />
i<br />
To make contact with the time-dependent configuration probability, we<br />
<strong>in</strong>troduce the formal state vector<br />
|Φ(t)〉 = ∑ {n i}<br />
P ({n i }; t) |{n i }〉 , (7.158)<br />
whereupon the l<strong>in</strong>ear time evolution accord<strong>in</strong>g to the master equation is translated<br />
<strong>in</strong>to an “imag<strong>in</strong>ary-time” Schröd<strong>in</strong>ger equation<br />
∂|Φ(t)〉<br />
∂t<br />
= −H |Φ(t)〉 , |Φ(t)〉 =e −Ht |Φ(0)〉 . (7.159)<br />
The stochastic quasi-Hamiltonian (rather, the time evolution or Liouville operator)<br />
for the on-site reaction processes is a sum of local terms, H reac =<br />
∑<br />
i H i(a † i ,a i); e.g., for the b<strong>in</strong>ary annihilation and coagulation reactions,<br />
(<br />
H i (a † i ,a i)=−λ 1 − a † 2 )<br />
i a 2 i − λ ′( )<br />
1 − a † i a † i a2 i . (7.160)<br />
The two contributions for each process may be physically <strong>in</strong>terpreted as follows:<br />
The first term corresponds to the actual process under consideration,<br />
and describes how many particles are annihilated and (re-)created <strong>in</strong> each<br />
reaction. The second term gives the “order” of each reaction, i.e., the number<br />
operator a † i a i appears to the kth power, but <strong>in</strong> normal-ordered form as<br />
a † k<br />
i a<br />
k<br />
i ,forakth-order process. Note that the reaction Hamiltonians such as<br />
(7.160) are non-Hermitean, reflect<strong>in</strong>g the particle creations and destructions.<br />
In a similar manner, hopp<strong>in</strong>g between neighbour<strong>in</strong>g sites 〈ij〉 is represented<br />
<strong>in</strong> this formalism through<br />
(<br />
)<br />
a † i − a† j<br />
)(a i − a j . (7.161)<br />
H diff = D ∑ 〈ij〉<br />
Our goal is of course to compute averages with respect to the configurational<br />
probability distribution P ({n i }; t); this is achieved by means of the<br />
projection state 〈P| = 〈0| ∏ i eai , which satisfies 〈P|0〉 =1and〈P|a † i = 〈P|,
332 U. C. Täuber<br />
s<strong>in</strong>ce [e ai ,a † j ]=eai δ ij . For the desired statistical averages of observables that<br />
must be expressible <strong>in</strong> terms of the occupation numbers {n i }, we then obta<strong>in</strong><br />
〈F (t)〉 = ∑ {n i}<br />
F ({n i }) P ({n i }; t) =〈P| F ({a † i a i}) |Φ(t)〉 . (7.162)<br />
Let first us explore the consequences of probability conservation, i.e., 1 =<br />
〈P|Φ(t)〉 = 〈P|e −Ht |Φ(0)〉. This requires 〈P|H = 0; upon commut<strong>in</strong>g e ∑ i ai<br />
with H, effectively the creation operators become shifted a † i → 1+a† i , whence<br />
this condition is fulfilled provided H i (a † i → 1,a i) = 0, which is <strong>in</strong>deed satisfied<br />
by our explicit expressions (7.160) and (7.161). By this prescription, we may<br />
also <strong>in</strong> averages replace a † i a i → a i , i.e., the particle density becomes a(t) =<br />
〈a i 〉, and the two-po<strong>in</strong>t operator a † i a i a † j a j → a i δ ij + a i a j .<br />
In the bosonic operator representation above, we have assumed that there<br />
exist no restrictions on the particle occupation numbers n i on each site. If,<br />
however, there is a maximum n i ≤ 2s+1, one may <strong>in</strong>stead employ a representation<br />
<strong>in</strong> terms of sp<strong>in</strong> s operators. For example, particle exclusion systems<br />
with n i = 0 or 1 can thus be mapped onto non-Hermitean sp<strong>in</strong> 1/2 “quantum”<br />
systems. Specifically <strong>in</strong> one dimension, such representations <strong>in</strong> terms of <strong>in</strong>tegrable<br />
sp<strong>in</strong> cha<strong>in</strong>s have proved a fruitful tool; for overviews, see Refs. [51–54].<br />
An alternative approach uses the bosonic theory, but encodes the site occupation<br />
restrictions through appropriate exponentials <strong>in</strong> the number operators<br />
e −a† i ai [55].<br />
We may now follow an established route <strong>in</strong> quantum many-particle theory<br />
[56] and proceed towards a field theory representation through construct<strong>in</strong>g<br />
the path <strong>in</strong>tegral equivalent to the “Schröd<strong>in</strong>ger” dynamics (7.159) based<br />
on coherent states, which are right eigenstates of the annihilation operator,<br />
a i |φ i 〉 = φ i |φ i 〉, with complex eigenvalues φ i . Explicitly, one f<strong>in</strong>ds<br />
(<br />
|φ i 〉 =exp − 1 )<br />
2 |φ i| 2 + φ i a † i |0〉 , (7.163)<br />
satisfy<strong>in</strong>g the overlap and (over-)completeness relations<br />
〈φ j |φ i 〉 =exp<br />
(− 1 2 |φ i| 2 − 1 ) ∫ ∏<br />
2 |φ j| 2 + φ ∗ d 2 φ i<br />
j φ i ,<br />
π |{φ i}〉 〈{φ i }| =1.<br />
i<br />
(7.164)<br />
Upon splitt<strong>in</strong>g the temporal evolution (7.159) <strong>in</strong>to <strong>in</strong>f<strong>in</strong>itesimal steps, and<br />
<strong>in</strong>sert<strong>in</strong>g Eq. (7.164) at each time step, standard procedures (elaborated <strong>in</strong><br />
detail <strong>in</strong> Ref. [12]) yield eventually<br />
∫ ∏<br />
〈F (t)〉 ∝<br />
i<br />
D[φ i ] D[φ ∗ i ] F ({φ i })e −A[φ∗ i ,φi] , (7.165)<br />
with the action
A[φ ∗ i ,φ i ]= ∑ i<br />
7 Field-Theory Approaches to Nonequilibrium Dynamics 333<br />
( ∫ tf<br />
[<br />
] )<br />
−φ i (t f )+ dt φ ∗ ∂φ i<br />
i<br />
0 ∂t + H i(φ ∗ i ,φ i ) −¯n 0 φ ∗ i (0) , (7.166)<br />
where the first term orig<strong>in</strong>ates from the projection state, and the last one from<br />
the <strong>in</strong>itial Poisson distribution. Notice that <strong>in</strong> the Hamiltonian, the creation<br />
and annihilation operators a † i and a i are simply replaced with the complex<br />
numbers φ ∗ i and φ i, respectively.<br />
Tak<strong>in</strong>g the cont<strong>in</strong>uum limit, φ i (t) → ψ(x,t), φ ∗ i (t) → ˆψ(x,t), the “bulk”<br />
part of the action becomes<br />
∫ ∫ [ ( )<br />
]<br />
A[ ˆψ, ∂<br />
ψ] = d d x dt ˆψ<br />
∂t − D ∇2 ψ + H reac ( ˆψ, ψ) , (7.167)<br />
where the hopp<strong>in</strong>g term (7.161) has naturally turned <strong>in</strong>to a diffusion propagator.<br />
We have thus arrived at a microscopic stochastic field theory for reaction–<br />
diffusion processes, with no assumptions whatsoever on the form of the (<strong>in</strong>ternal)<br />
noise. This is a crucial <strong>in</strong>gredient for nonequilibrium dynamics, and<br />
we may now use Eq. (7.167) as a basis for systematic coarse-gra<strong>in</strong><strong>in</strong>g and the<br />
renormalisation group analysis. Return<strong>in</strong>g to our example of pair annihilation<br />
and coagulation, the reaction part of the action (7.167) reads<br />
H reac ( ˆψ,<br />
(<br />
ψ) =−λ 1 − ˆψ 2) ( ˆψ)<br />
ψ 2 − λ ′ 1 − ˆψψ 2 , (7.168)<br />
see Eq. (7.160). Let us have a look at the classical field equations, namely<br />
δA/δψ = 0, which is always solved by ˆψ = 1, reflect<strong>in</strong>g probability conservation,<br />
and δA/δ ˆψ = 0, which, upon <strong>in</strong>sert<strong>in</strong>g ˆψ = 1 gives here<br />
∂ψ(x,t)<br />
= D ∇ 2 ψ(x,t) − (2λ + λ ′ ) ψ(x,t) 2 , (7.169)<br />
∂t<br />
i.e., essentially the mean-field rate equation for the local particle density<br />
ψ(x,t), see Eq. (7.149), supplemented with diffusion. The field theory action<br />
(7.167), derived from the master equation (7.154), then provides a means<br />
of <strong>in</strong>clud<strong>in</strong>g fluctuations <strong>in</strong> our analysis.<br />
Before we proceed with this program, it is <strong>in</strong>structive to perform a shift<br />
<strong>in</strong> the field ˆψ about the mean-field solution, ˆψ(x, t) =1+˜ψ(x, t), whereupon<br />
the reaction Hamiltonian density (7.168) becomes<br />
H reac ( ˜ψ, ψ) =(2λ + λ ′ ) ˜ψψ 2 +(λ + λ ′ ) ˜ψ 2 ψ 2 . (7.170)<br />
In addition to the diffusion propagator, the annihilation and coagulation<br />
processes thus give identical three- and four-po<strong>in</strong>t vertices; aside from nonuniversal<br />
amplitudes, one should therefore obta<strong>in</strong> identical scal<strong>in</strong>g behaviour<br />
for both b<strong>in</strong>ary reactions <strong>in</strong> the asymptotic long-time limit [57]. Lastly, we<br />
remark that if we <strong>in</strong>terpret the action A[ ˜ψ, ψ] as a response functional (7.34),<br />
despite the fields ˜ψ not be<strong>in</strong>g purely imag<strong>in</strong>ary, our field theory becomes formally<br />
equivalent to a “Langev<strong>in</strong>” equation, where<strong>in</strong> additive noise is added to
334 U. C. Täuber<br />
Eq. (7.169), albeit with negative correlator L[ψ] =−(λ + λ ′ ) ψ 2 , which represents<br />
“imag<strong>in</strong>ary” multiplicative noise. This Langev<strong>in</strong> description is thus not<br />
well-def<strong>in</strong>ed; however, one may render the noise correlator positive through<br />
a nonl<strong>in</strong>ear Cole–Hopf transformation ˜ψ =e˜ρ , ψ = ρ e −˜ρ such that ˜ψψ = ρ,<br />
with Jacobian 1, but at the expense of “diffusion noise” ∝ Dρ(∇˜ρ ) 2 <strong>in</strong> the action<br />
[58]. In summary, b<strong>in</strong>ary (and higher-order) annihilation and coagulation<br />
processes cannot be cast <strong>in</strong>to a Langev<strong>in</strong> framework <strong>in</strong> any simple manner.<br />
7.2.3 Diffusion-Limited S<strong>in</strong>gle-Species Annihilation Processes<br />
We beg<strong>in</strong> by analys<strong>in</strong>g diffusion-limited s<strong>in</strong>gle-species annihilation kA →∅<br />
[57, 59]. The correspond<strong>in</strong>g field theory action (7.167) reads<br />
∫ ∫ [ ( )<br />
A[ ˆψ, ∂<br />
ψ] = d d x dt ˆψ<br />
∂t − D∇2 ψ − λ<br />
(1 − ˆψ k) ]<br />
ψ k , (7.171)<br />
which for k ≥ 3 allows no (obvious) equivalent Langev<strong>in</strong> description. Straightforward<br />
power count<strong>in</strong>g gives the scal<strong>in</strong>g dimension for the annihilation rate,<br />
[λ] =µ 2−(k−1)d , which suggests the upper critical dimension d c (k) =2/(k−1).<br />
Thus we expect mean-field behaviour ∼ (λt) −1/(k−1) , see Eq. (7.151), <strong>in</strong> any<br />
physical dimension for k>3, logarithmic corrections at d c =1fork =3and<br />
at d c =2fork = 2, and nonclassical power laws for pair annihilation only <strong>in</strong><br />
one dimension. The field theory def<strong>in</strong>ed by the action (7.171) has two vertices,<br />
the “annihilation” s<strong>in</strong>k with k <strong>in</strong>com<strong>in</strong>g l<strong>in</strong>es only, and the “scatter<strong>in</strong>g” vertex<br />
with k <strong>in</strong>com<strong>in</strong>g and k outgo<strong>in</strong>g l<strong>in</strong>es. Neither allows for propagator renormalisation,<br />
hence the model rema<strong>in</strong>s massless with exact scal<strong>in</strong>g exponents<br />
η =0andz = 2, i.e., diffusive dynamics.<br />
In addition, the entire perturbation expansion for the renormalisation for<br />
the annihilation vertices is merely a geometric series of the one-loop diagram,<br />
see Fig. 7.6 for the pair annihilation case (k = 2). If we def<strong>in</strong>e the renormalised<br />
effective coupl<strong>in</strong>g accord<strong>in</strong>g to<br />
g R = Z g<br />
λ<br />
D B kd µ −2(1−d/dc) , (7.172)<br />
where B kd = k! Γ (2−d/d c ) d c /k d/2 (4π) d/dc , we obta<strong>in</strong> for the s<strong>in</strong>gle nontrivial<br />
renormalisation constant<br />
+ + +<br />
. . .<br />
+<br />
. . .<br />
+<br />
. . .<br />
Fig. 7.6. Vertex renormalisation for diffusion-limited b<strong>in</strong>ary annihilation A+A →∅
7 Field-Theory Approaches to Nonequilibrium Dynamics 335<br />
Z −1<br />
g<br />
=1+ λB kd µ −2(1−d/dc)<br />
D (d c − d)<br />
(7.173)<br />
to all orders. Consequently, the associated RG beta function becomes<br />
β g = µ ∂<br />
∂µ ∣ g R = − 2 g R<br />
(d − d c + g R ) , (7.174)<br />
0<br />
d c<br />
with the Gaussian fixed po<strong>in</strong>t g ∗ 0 = 0 stable for d > d c (k) = 2/(k − 1),<br />
lead<strong>in</strong>g to the mean-field power laws (7.151), whereas for d
336 U. C. Täuber<br />
difference of particle numbers (even locally), i.e., there is a conservation law<br />
for c(t) =a(t) − b(t) =c(0) [61]. The rate equations for the concentrations<br />
ȧ(t) =−λa(t) b(t) =ḃ(t) (7.181)<br />
are for equal <strong>in</strong>itial densities a(0) = b(0) solved by the s<strong>in</strong>gle-species pair<br />
annihilation mean-field power law<br />
a(t) =b(t) ∼ (λt) −1 , (7.182)<br />
whereas for unequal <strong>in</strong>itial densities c(0) = a(0) − b(0) > 0, say, the majority<br />
species a(t) → a ∞ = c(0) > 0ast →∞, and the m<strong>in</strong>ority population disappears,<br />
b(t) → 0. From Eq. (7.181) we obta<strong>in</strong> for d>d c = 2 the exponential<br />
approach<br />
a(t) − a ∞ ∼ b(t) ∼ e −c(0) λt . (7.183)<br />
Mapp<strong>in</strong>g the associated master equation onto a cont<strong>in</strong>uum field theory<br />
(7.167), the reaction term now reads (with the fields ψ and ϕ represent<strong>in</strong>g the<br />
A and B particles, respectively) [62]<br />
H reac ( ˆψ,<br />
(<br />
ψ, ˆϕ, ϕ) =−λ 1 − ˆψ<br />
)<br />
ˆϕ ψϕ . (7.184)<br />
As <strong>in</strong> the s<strong>in</strong>gle-species case, there is no propagator renormalisation, and<br />
moreover the Feynman diagrams for the renormalised reaction vertex are of<br />
precisely the same form as for A + A →∅, see Fig. 7.6. Thus, for unequal<br />
<strong>in</strong>itial densities, c(0) > 0, the mean-field power law ∼ λt <strong>in</strong> the exponent of<br />
Eq. (7.183) becomes aga<strong>in</strong> replaced with (Dt) d/2 <strong>in</strong> dimensions d ≤ d c =2,<br />
lead<strong>in</strong>g to stretched exponential time dependence,<br />
d
7 Field-Theory Approaches to Nonequilibrium Dynamics 337<br />
∫<br />
c(x,t)= d d x ′ G 0 (x − x ′ ,t) c(x ′ , 0) . (7.188)<br />
Let us furthermore assume a Poisson distribution for the <strong>in</strong>itial density correlations<br />
(<strong>in</strong>dicated by an overbar), a(x, 0) a(x ′ , 0) = a(0) 2 + a(0) δ(x − x ′ )=<br />
b(x, 0) b(x ′ , 0) and a(x, 0) b(x ′ (0) = a(0) 2 , which implies c(x, 0) c(x ′ , 0) =<br />
2 a(0)δ(x−x ′ ). Averag<strong>in</strong>g over the <strong>in</strong>itial conditions then yields with Eq. (7.188)<br />
∫<br />
c(x,t) 2 =2a(0) d d x ′ G 0 (x − x ′ ,t) 2 =2a(0) (8πDt) −d/2 ; (7.189)<br />
s<strong>in</strong>ce the distribution for c will be a Gaussian, we thus obta<strong>in</strong> for the local<br />
density excess orig<strong>in</strong>at<strong>in</strong>g <strong>in</strong> a random <strong>in</strong>itial fluctuation,<br />
√ √<br />
2 a(0)<br />
|c(x,t)| =<br />
π c(x,t)2 =2<br />
π (8πDt)−d/4 . (7.190)<br />
In dimensions d2and∼t −d/2 for d
338 U. C. Täuber<br />
and D + A →∅. We may then obviously identify A = C and B = D, which<br />
leads back to the case of two-species pair annihilation. Generally, with<strong>in</strong> essentially<br />
mean-field theory one f<strong>in</strong>ds for cyclic multi-species annihilation processes<br />
a i (t) ∼ t −α(q,d) , where for<br />
{<br />
4 q =2, 4, 6,...<br />
2
7 Field-Theory Approaches to Nonequilibrium Dynamics 339<br />
[λ] =µ 2−d ,isthusirrelevant <strong>in</strong> the RG sense, and can be dropped for the<br />
computation of universal, asymptotic scal<strong>in</strong>g properties.<br />
The action (7.195) with λ =0isknownasReggeon field theory [65], and<br />
its basic characteristic is its <strong>in</strong>variance under rapidity <strong>in</strong>version S(x,t) ↔<br />
− ˜S(x, −t). If we <strong>in</strong>terpret Eq. (7.195) as a response functional, we see that it<br />
becomes formally equivalent to a stochastic process with multiplicative noise<br />
(〈ζ(x,t)〉 = 0) captured by the Langev<strong>in</strong> equation [66, 67]<br />
∂S(x,t)<br />
= −D ( r − ∇ 2) S(x,t) − uS(x,t) 2 + ζ(x,t) ,<br />
∂t<br />
(7.196)<br />
〈ζ(x,t) ζ(x ′ ,t ′ )〉 =2uS(x,t) δ(x − x ′ ) δ(t − t ′ ) (7.197)<br />
(for a more accurate mapp<strong>in</strong>g procedure, see Ref. [68]), which is essentially a<br />
noisy Fisher–Kolmogorov equation (7.193), with the noise correlator (7.197)<br />
ensur<strong>in</strong>g that the fluctuations <strong>in</strong>deed cease <strong>in</strong> the absorb<strong>in</strong>g state where<br />
〈S〉 = 0. It has moreover been established [69–71] that the action (7.195) describes<br />
the scal<strong>in</strong>g properties of critical directed percolation (DP) clusters [72],<br />
illustrated <strong>in</strong> Fig. 7.7. Indeed, if the DP growth direction is labeled as “time”<br />
t, we see that the structure of the DP clusters emerges from the basic decay,<br />
branch<strong>in</strong>g, and coagulation reactions encoded <strong>in</strong> Eq. (7.194).<br />
t ⃗<br />
Fig. 7.7. Directed percolation process (left) and critical DP cluster (right)<br />
In fact, the field theory action should govern the scal<strong>in</strong>g properties of<br />
generic cont<strong>in</strong>uous nonequilibrium phase transitions from active to <strong>in</strong>active,<br />
absorb<strong>in</strong>g states, namely for an order parameter with Markovian stochastic<br />
dynamics that is decoupled from any other slow variable, and <strong>in</strong> the absence<br />
of quenched randomness [71, 73]. This DP conjecture follows from the follow<strong>in</strong>g<br />
phenomenological approach [68] to simple epidemic processes (SEP), or<br />
epidemics with recovery [46]:<br />
1. A “susceptible” medium becomes locally “<strong>in</strong>fected”, depend<strong>in</strong>g on the density<br />
n of neighbor<strong>in</strong>g “sick” <strong>in</strong>dividuals. The <strong>in</strong>fected regions recover after<br />
a brief time <strong>in</strong>terval.<br />
2. The state n =0isabsorb<strong>in</strong>g. It describes the ext<strong>in</strong>ction of the “disease’.<br />
3. The disease spreads out diffusively via the short-range <strong>in</strong>fection 1. of neighbor<strong>in</strong>g<br />
susceptible regions.
340 U. C. Täuber<br />
4. Microscopic fast degrees of freedom are <strong>in</strong>corporated as local noise or<br />
stochastic forces that respect statement 2., i.e., the noise alone cannot<br />
regenerate the disease.<br />
These <strong>in</strong>gredients are captured by the coarse-gra<strong>in</strong>ed mesoscopic Langev<strong>in</strong><br />
equation ∂ t n = D ( ∇ 2 − R[n] ) n + ζ with a reaction functional R[n], and<br />
s stochastic noise correlator of the form L[n] =nN[n]. Near the ext<strong>in</strong>ction<br />
threshold, we may expand R[n] =r + un + ..., N[n] =v + ..., and higherorder<br />
terms turn out to be irrelevant <strong>in</strong> the RG sense. Upon rescal<strong>in</strong>g, we<br />
recover the Reggeon field theory action (7.195) for DP as the correspond<strong>in</strong>g<br />
response functional (7.34).<br />
We now proceed to an explicit evaluation of the DP critical exponents<br />
to one-loop order, closely follow<strong>in</strong>g the recipes given <strong>in</strong> Subsects. 7.1.3–7.1.5.<br />
The lowest-order fluctuation contribution to the two-po<strong>in</strong>t vertex function<br />
Γ (1,1) (q,ω) (propagator self-energy) is depicted <strong>in</strong> Fig. 7.8(a). The Feynman<br />
rules of Subsect. 7.1.3 yield the correspond<strong>in</strong>g analytic expression<br />
Γ (1,1) (q,ω)=iω + D (r + q 2 )+ u2<br />
D<br />
∫<br />
k<br />
1<br />
iω/2D + r + q 2 /4+k 2 . (7.198)<br />
The criticality condition Γ (1,1) (0, 0) = 0 at r = r c<br />
fluctuation-<strong>in</strong>duced shift of the percolation threshold<br />
provides us with the<br />
∫ Λ<br />
r c = − u2 1<br />
D 2 r c + k 2 + O(u4 ) . (7.199)<br />
k<br />
Insert<strong>in</strong>g τ = r − r c <strong>in</strong>to Eq. (7.198), we then f<strong>in</strong>d to this order<br />
∫<br />
Γ (1,1) (q,ω)=iω + D (τ + q 2 ) − u2 iω/2D + τ + q 2 /4<br />
D k 2 (iω/2D + τ + q 2 /4+k 2 ) , (7.200)<br />
k<br />
and the diagram <strong>in</strong> Fig. 7.8(b) for the three-po<strong>in</strong>t vertex functions, evaluated<br />
at zero external wavevectors and frequencies, gives<br />
(<br />
)<br />
Γ (1,2) ({0}) =−Γ (2,1) ({0}) =−2u 1 − 2u2 1<br />
D<br />
∫k<br />
2 (τ + k 2 ) 2 . (7.201)<br />
For the renormalisation factors, we use aga<strong>in</strong> the conventions (7.56) and<br />
(7.58), but with<br />
(a)<br />
(b)<br />
Fig. 7.8. DP renormalisation: one-loop diagrams for the vertex functions (a) Γ (1,1)<br />
(propagator self-energy), and (b) Γ (1,2) = −Γ (2,1) (nonl<strong>in</strong>ear vertices)
7 Field-Theory Approaches to Nonequilibrium Dynamics 341<br />
u R = Z u uA 1/2<br />
d<br />
µ (d−4)/2 . (7.202)<br />
Because of rapidity <strong>in</strong>version <strong>in</strong>variance, Z ˜S<br />
= Z S . With Eq. (7.53) the derivatives<br />
of Γ (1,1) with respect to ω, q 2 ,andτ, as well as the one-loop result for<br />
Γ (1,2) <strong>in</strong> Eq. (7.201), all evaluated at the normalisation po<strong>in</strong>t τ R =1,provide<br />
us with the Z factors<br />
Z S =1− u2 A d µ −ɛ<br />
2D 2 ɛ<br />
Z τ =1− 3u2 A d µ −ɛ<br />
4D 2 ɛ<br />
From these we <strong>in</strong>fer the RG flow functions<br />
, Z D =1+ u2 A d µ −ɛ<br />
4D 2 ɛ<br />
, Z u =1− 5u2 A d µ −ɛ<br />
4D 2 ɛ<br />
,<br />
. (7.203)<br />
γ S = v R /2 , γ D = −v R /4 , γ τ = −2+3v R /4 , (7.204)<br />
with the renormalised effective coupl<strong>in</strong>g<br />
v R = Z2 u<br />
Z 2 D<br />
whose RG beta function is to this order<br />
u 2<br />
D 2 A d µ d−4 , (7.205)<br />
β v = v R<br />
[<br />
−ɛ +3vR + O(v 2 R) ] . (7.206)<br />
For d>d c = 4, the Gaussian fixed po<strong>in</strong>t v ∗ 0 = 0 is stable, and we recover<br />
the mean-field critical exponents. For ɛ =4− d>0, we f<strong>in</strong>d the nontrivial<br />
IR-stable RG fixed po<strong>in</strong>t<br />
v ∗ = ɛ/3+O(ɛ 2 ) . (7.207)<br />
Sett<strong>in</strong>g up and solv<strong>in</strong>g the RG equation (7.65) for the vertex function<br />
proceeds just as <strong>in</strong> Subsect. 7.1.5. With the identifications (7.83) (with a =0)<br />
we thus obta<strong>in</strong> the DP critical exponents to first order <strong>in</strong> ɛ,<br />
η = − ɛ 6 + O(ɛ2 ) ,<br />
1<br />
ν =2− ɛ 4 + O(ɛ2 ) ,<br />
z =2− ɛ<br />
12 + O(ɛ2 ) . (7.208)<br />
In the vic<strong>in</strong>ity of v ∗ , the solution of the RG equation for the order parameter<br />
reads, recall<strong>in</strong>g that [S] =µ d/2 ,<br />
)<br />
〈S R (τ R ,t)〉 ≈µ d/2 l (d−γ∗ S )/2 Ŝ<br />
(τ R l γ∗ τ ,v<br />
∗<br />
R ,D R µ 2 l 2+γ∗ D t , (7.209)<br />
which leads to the follow<strong>in</strong>g scal<strong>in</strong>g relations and explicit exponent values,<br />
β =<br />
ν(d + η)<br />
2<br />
=1− ɛ 6 + O(ɛ2 ) , α = β<br />
zν =1− ɛ 4 + O(ɛ2 ) . (7.210)<br />
The scal<strong>in</strong>g exponents for critical directed percolation are known analytically<br />
for a plethora of physical quantities (but the reader should beware that
342 U. C. Täuber<br />
various different conventions are used <strong>in</strong> the literature); for the two-loop results<br />
to order ɛ 2 <strong>in</strong> the perturbative dimensional expansion, see Ref. [68]. In<br />
Table 7.2, we compare the O(ɛ) values with the results from Monte Carlo<br />
computer simulations, which allow the DP critical exponents to be measured<br />
to high precision (for recent overviews on simulation results for DP and other<br />
absorb<strong>in</strong>g state phase transitions, see Refs. [74, 75]). Yet unfortunately, there<br />
are to date hardly any real experiments that would confirm the DP conjecture<br />
[71, 73] and actually measure the scal<strong>in</strong>g exponents for this prom<strong>in</strong>ent<br />
nonequilibrium universality class.<br />
Table 7.2. Comparison of the DP critical exponent values from Monte Carlo simulations<br />
with the results from the ɛ expansion<br />
Scal<strong>in</strong>g exponent d =1 d =2 d =4− ɛ<br />
ξ ∼|τ| −ν ν ≈ 1.100 ν ≈ 0.735 ν =1/2+ɛ/16 + O(ɛ 2 )<br />
t c ∼ ξ z ∼|τ| −zν z ≈ 1.576 z ≈ 1.73 z =2− ɛ/12 + O(ɛ 2 )<br />
a ∞ ∼|τ| β β ≈ 0.2765 β ≈ 0.584 β =1− ɛ/6+O(ɛ 2 )<br />
a c(t) ∼ t −α α ≈ 0.160 α ≈ 0.46 α =1− ɛ/4+O(ɛ 2 )<br />
7.2.6 Dynamic Isotropic Percolation and Multi-Species Variants<br />
An <strong>in</strong>terest<strong>in</strong>g variant of active to absorb<strong>in</strong>g state phase transitions emerges<br />
when we modify the SEP rules (1) and (2) <strong>in</strong> Subsect. 7.2.5 to<br />
1. The susceptible medium becomes <strong>in</strong>fected, depend<strong>in</strong>g on the densities n<br />
and m of sick <strong>in</strong>dividuals and the “debris”, respectively. After a brief time<br />
<strong>in</strong>terval, the sick <strong>in</strong>dividuals decay <strong>in</strong>to immune debris, which ultimately<br />
stops the disease locally by exhaust<strong>in</strong>g the supply of susceptible regions.<br />
2. The states with n = 0 and any spatial distribution of m are absorb<strong>in</strong>g,<br />
and describe the ext<strong>in</strong>ction of the disease.<br />
Here, the debris is given by the accumulated decay products,<br />
∫ t<br />
m(x,t)=κ n(x,t ′ )dt ′ . (7.211)<br />
−∞<br />
After rescal<strong>in</strong>g, this general epidemic process (GEP) or epidemic with removal<br />
[46] is described <strong>in</strong> terms of the mesoscopic Langev<strong>in</strong> equation [76]<br />
∂S(x,t)<br />
= −D ( r − ∇ 2) ∫ t<br />
S(x,t) − DuS(x,t) S(x,t ′ ) Dt ′ + ζ(x,t) ,<br />
∂t<br />
−∞<br />
(7.212)<br />
with noise correlator (7.197). The associated response functional reads [77,78]
7 Field-Theory Approaches to Nonequilibrium Dynamics 343<br />
∫ ∫ [ (<br />
A[ ˜S,S]=<br />
∂<br />
d d x dt ˜S<br />
∂t + D ( r − ∇ 2)) t<br />
]<br />
S − u ˜S 2 S + DuS∫<br />
S(t ′ ) .<br />
(7.213)<br />
For the field theory thus def<strong>in</strong>ed, one may take the quasistatic limit by<br />
<strong>in</strong>troduc<strong>in</strong>g the fields<br />
˜ϕ(x) = ˜S(x,t→∞) , ϕ(x) =D<br />
∫ ∞<br />
−∞<br />
S(x,t ′ )dt ′ . (7.214)<br />
For t →∞, the action (7.213) thus becomes<br />
∫ [<br />
]<br />
A qst [˜ϕ, ϕ] = d d x ˜ϕ r − ∇ 2 − u (˜ϕ − ϕ) ϕ, (7.215)<br />
which is known to describe the critical exponents of isotropic percolation [79].<br />
An isotropic percolation cluster is shown <strong>in</strong> Fig. 7.9, to be contrasted with<br />
the anisotropic scal<strong>in</strong>g evident <strong>in</strong> Fig. 7.7(b). The upper critical dimension of<br />
isotropic percolation is d c = 6, and an explicit calculation, with the diagrams<br />
of Fig. 7.8, but <strong>in</strong>volv<strong>in</strong>g the static propagators G 0 (q) =1/(r +q 2 ), yields the<br />
follow<strong>in</strong>g critical exponents for isotropic percolation, to first order <strong>in</strong> ɛ =6−d,<br />
η = − ɛ<br />
21 + O(ɛ2 ) ,<br />
1<br />
ν =2− 5 ɛ<br />
21 + O(ɛ2 ) , β =1− ɛ 7 + O(ɛ2 ) . (7.216)<br />
In order to calculate the dynamic critical exponent for this dynamic isotropic<br />
percolation (dIP) universality class, we must return to the full action (7.213).<br />
Once aga<strong>in</strong>, with the diagrams of Fig. 7.8, but now <strong>in</strong>volv<strong>in</strong>g a temporally<br />
nonlocal three-po<strong>in</strong>t vertex, one then arrives at<br />
z =2− ɛ 6 + O(ɛ2 ) . (7.217)<br />
For a variety of two-loop results, the reader is referred to Ref. [68]. It is also<br />
possible to describe the crossover from isotropic to directed percolation with<strong>in</strong><br />
this field-theoretic framework [80, 81].<br />
Fig. 7.9. Isotropic percolation cluster
344 U. C. Täuber<br />
Let us next consider multi-species variants of directed percolation processes,<br />
which can be obta<strong>in</strong>ed <strong>in</strong> the particle language by coupl<strong>in</strong>g the DP reactions<br />
A i →∅, A i ⇋ A i + A i via processes of the form A i ⇋ A j + A j (with j ≠ i);<br />
or directly by the correspond<strong>in</strong>g generalisation with<strong>in</strong> the Langev<strong>in</strong> representation<br />
with 〈ζ i (x,t)〉 =0,<br />
∂S i<br />
∂t = D (<br />
i ∇ 2 − R i [S i ] ) S i + ζ i , R i [S i ]=r i + ∑ j<br />
g ij S j + ... , (7.218)<br />
〈ζ i (x,t)ζ j (x ′ ,t ′ )〉=2S i N i [S i ] δ(x − x ′ ) δ(t − t ′ ) δ ij ,N i [S i ]=u i + ... (7.219)<br />
The ensu<strong>in</strong>g renormalisation factors turn out to be precisely as for s<strong>in</strong>glespecies<br />
DP, and consequently the generical critical behaviour even <strong>in</strong> such<br />
multi-species systems is governed by the DP universality class [58]. For example,<br />
the predator ext<strong>in</strong>ction threshold for the stochastic Lotka–Volterra<br />
system mentioned <strong>in</strong> Subsect. 7.2.1 is characterised by the DP exponents as<br />
well [47]. But these reactions also generate A i → A j , caus<strong>in</strong>g additional terms<br />
∑<br />
j≠i g j S j <strong>in</strong> Eq. (7.218). Asymptotically, the <strong>in</strong>ter-species coupl<strong>in</strong>gs become<br />
unidirectional, which allows for the appearance of special multicritical po<strong>in</strong>ts<br />
when several r i = 0 simultaneously [82]. This leads to a hierarchy of order<br />
parameter exponents β k on the kth level of a unidirectional cascade, with<br />
β 1 =1− ɛ 6 + O(ɛ2 ) ,<br />
β 2 = 1 2 − 13 ɛ<br />
96 + O(ɛ2 ) ,..., β k = 1 − O(ɛ) ; (7.220)<br />
2k for the associated crossover exponent, one can show Φ =1toall orders [58].<br />
Quite analogous features emerge for multi-species dIP processes [58, 68].<br />
7.2.7 Conclud<strong>in</strong>g Remarks<br />
In these lecture notes, I have described how stochastic processes can be<br />
mapped onto field theory representations, start<strong>in</strong>g either from a mesoscopic<br />
Langev<strong>in</strong> equation for the coarse-gra<strong>in</strong>ed densities of the relevant order parameter<br />
fields and conserved quantities, or from a more microsopic master<br />
equation for <strong>in</strong>teract<strong>in</strong>g particle systems. The dynamic renormalisation group<br />
method can then be employed to study and characterise the universal scal<strong>in</strong>g<br />
behaviour near cont<strong>in</strong>uous phase transitions both <strong>in</strong> and far from thermal<br />
equilibrium, and for systems that generically display scale-<strong>in</strong>variant features.<br />
While the critical dynamics near equilibrium phase transitions has been thoroughly<br />
<strong>in</strong>vestigated experimentally <strong>in</strong> the past three decades, regrettably such<br />
direct experimental verification of the by now considerable amount of theoretical<br />
work on nonequilibrium systems is largely amiss. In this respect, applications<br />
of the expertise ga<strong>in</strong>ed <strong>in</strong> the nonequilibrium statistical mechanics<br />
of complex cooperative behaviour to biological systems might prove fruitful<br />
and constitutes a promis<strong>in</strong>g venture. One must bear <strong>in</strong> m<strong>in</strong>d, however, that<br />
nonuniversal features are often crucial for the relevant questions <strong>in</strong> biology.
7 Field-Theory Approaches to Nonequilibrium Dynamics 345<br />
In part, the lack of clearcut experimental evidence may be due to the fact<br />
that asymptotic universal properties are perhaps less prom<strong>in</strong>ent <strong>in</strong> accessible<br />
nonequilibrium systems, ow<strong>in</strong>g to long crossover times. Yet fluctuations do<br />
tend to play a more important role <strong>in</strong> systems that are driven away from thermal<br />
equilibrium, and the concept of universality classes, despite the undoubtedly<br />
much <strong>in</strong>creased richness <strong>in</strong> dynamical systems, should still be useful. For<br />
example, we have seen that the directed percolation universality class quite<br />
generically describes the critical properties of phase transitions from active<br />
to <strong>in</strong>active, absorb<strong>in</strong>g states, which abound <strong>in</strong> nature. The few exceptions<br />
to this rule either require the coupl<strong>in</strong>g to another conserved mode [83, 84];<br />
the presence, on a mesoscopic level, of additional symmetries that preclude<br />
the spontaneous decay A → ∅ as <strong>in</strong> the so-called parity-conserv<strong>in</strong>g (PC)<br />
universality class, represented by branch<strong>in</strong>g and annihilat<strong>in</strong>g random walks<br />
A → (n +1)A with n even, andA + A →∅[85] (for recent developments<br />
based on nonperturbative RG approaches, see Ref. [86]); or the absence of any<br />
first-order reactions, as <strong>in</strong> the (by now rather notorious) pair contact process<br />
with diffusion (PCPD) [87], which has so far eluded a successful field-theoretic<br />
treatment [88]. A possible explanation for the fact that DP exponents have not<br />
been measured ubiquitously (yet) could be the <strong>in</strong>stability towards quenched<br />
disorder <strong>in</strong> the reaction rates [89].<br />
In reaction–diffusion systems, a complete classification of the scal<strong>in</strong>g properties<br />
<strong>in</strong> multi-species systems rema<strong>in</strong>s <strong>in</strong>complete, aside from pair annihilation<br />
and DP-like processes, and still constitutes a quite formidable program<br />
(for a recent overview over the present situation from a field-theoretic viewpo<strong>in</strong>t,<br />
see Ref. [12]). This is even more evident for nonequilibrium systems <strong>in</strong><br />
general, even when ma<strong>in</strong>ta<strong>in</strong>ed <strong>in</strong> driven steady states. Field-theoretic methods<br />
and the dynamic renormalisation group represent powerful tools that I<br />
believe will cont<strong>in</strong>ue to crucially complement exact solutions (usually of onedimensional<br />
models), other approximative approaches, and computer simulations,<br />
<strong>in</strong> our quest to further elucidate the <strong>in</strong>trigu<strong>in</strong>g cooperative behaviour<br />
of strongly <strong>in</strong>teract<strong>in</strong>g and fluctuat<strong>in</strong>g many-particle systems.<br />
Acknowledgements<br />
This work has been supported <strong>in</strong> part by the U.S. National Science Foundation<br />
through grant NSF DMR-0308548. I am <strong>in</strong>debted to many colleagues,<br />
students, and friends, from and with whom I had the pleasure to learn and<br />
research the material presented here; specifically I would like to mention<br />
Vamsi Akk<strong>in</strong>eni, Michael Bulenda, John Cardy, Olivier Deloubrière, Daniel<br />
Fisher, Re<strong>in</strong>hard Folk, Erw<strong>in</strong> Frey, Ivan Georgiev, Yad<strong>in</strong> Goldschmidt, Peter<br />
Grassberger, Henk Hilhorst, Haye H<strong>in</strong>richsen, Mart<strong>in</strong> Howard, Terry Hwa,<br />
Hannes Janssen, Bernhard Kaufmann, Mauro Mobilia, David Nelson, Beth<br />
Reid, Zoltán Rácz, Jaime Santos, Beate Schmittmann, Franz Schwabl, Steffen<br />
Trimper, Ben Vollmayr-Lee, Mark Washenberger, Fred van Wijland, and<br />
Royce Zia. Lastly, I would like to thank the organisers of the very enjoyable
346 U. C. Täuber<br />
Luxembourg Summer School on Age<strong>in</strong>g and the Glass Transition for their<br />
k<strong>in</strong>d <strong>in</strong>vitation, and my colleagues at the Laboratoire de Physique Théorique,<br />
Université de Paris-Sud Orsay, France and at the Rudolf Peierls Centre for<br />
Theoretical <strong>Physics</strong>, University of Oxford, U.K., where these lecture notes<br />
were conceived and written, for their warm hospitality.<br />
References<br />
1. P. Ramond: Field theory – a modern primer, (Benjam<strong>in</strong>/Cumm<strong>in</strong>gs, Read<strong>in</strong>g<br />
1981)<br />
2. D.J. Amit: Field theory, the renormalization group, and critical phenomena<br />
(World Scientific, S<strong>in</strong>gapore 1984)<br />
3. C. Itzykson and J.M. Drouffe: Statistical field theory (Cambridge University<br />
Press, Cambridge 1989)<br />
4. M. Le Bellac: Quantum and statistical field theory, (Oxford University Press,<br />
Oxford 1991)<br />
5. J. Z<strong>in</strong>n-Just<strong>in</strong>: Quantum field theory and critical phenomena (Clarendon Press,<br />
Oxford 1993)<br />
6. J. Cardy: Scal<strong>in</strong>g and renormalization <strong>in</strong> statistical physics (Cambridge University<br />
Press, Cambridge 1996)<br />
7. P.C. Hohenberg and B.I. Halper<strong>in</strong>: Rev. Mod. Phys. 49, 435 (1977)<br />
8. H.K. Janssen: Field-theoretic methods applied to critical dynamics. In: Dynamical<br />
critical phenomena and related topics, <strong>Lecture</strong> <strong>Notes</strong> <strong>in</strong> <strong>Physics</strong>, vol. 104,<br />
ed by C.P. Enz (<strong>Spr<strong>in</strong>ger</strong>, Heidelberg 1979), pp. 26–47<br />
9. R. Bausch, H.K. Janssen, and H. Wagner: Z. Phys. B24, 113 (1976)<br />
10. J.L. Cardy: Renormalisation group approach to reaction-diffusion problems. In:<br />
Proceed<strong>in</strong>gs of Mathematical Beauty of <strong>Physics</strong>, edbyJ.-B.Zuber,Adv.Ser.<br />
<strong>in</strong> Math. Phys. 24, 113 (1997)<br />
11. D.C. Mattis and M.L. Glasser: Rev. Mod. Phys. 70, 979 (1998)<br />
12. U.C. Täuber, M.J. Howard, and B.P. Vollmayr-Lee: J. Phys. A: Math. Gen.<br />
38, R79 (2005)<br />
13. U.C. Täuber: Critical dynamics: a field theory approach to equilibrium and<br />
non-equilibrium scal<strong>in</strong>g behavior, <strong>in</strong> preparation (to be published at Cambridge<br />
University Press, Cambridge); for completed chapters, see:<br />
http://www.phys.vt.edu/˜tauber/utaeuber.html<br />
14. R.A. Ferrell, N. Menyhàrd, H. Schmidt, F. Schwabl, and P. Szépfalusy: Phys.<br />
Rev. Lett. 18, 891 (1967); Ann. of Phys. 47, 565 (1968)<br />
15. B.I. Halper<strong>in</strong> and P.C. Hohenberg: Phys. Rev. 177, 952 (1969)<br />
16. B.I. Halper<strong>in</strong>, P.C. Hohenberg, and S.-k. Ma: Phys. Rev. Lett. 29, 1548 (1972)<br />
17. C. De Dom<strong>in</strong>icis, E. Bréz<strong>in</strong>, and J. Z<strong>in</strong>n-Just<strong>in</strong>: Phys. Rev. B12, 4945 (1975)<br />
18. H.K. Janssen: Z. Phys. B23, 377 (1976)<br />
19. C. De Dom<strong>in</strong>icis: J. Physique Colloque 37, C2247 (1976)<br />
20. P.C. Mart<strong>in</strong>, E.D. Siggia, and H.A. Rose: Phys. Rev. A8, 423 (1973)<br />
21. H. Wagner: Z. Phys. 195, 273 (1966)<br />
22. N.D. Merm<strong>in</strong> and H. Wagner: Phys. Rev. Lett. 17, 1133 (1966)<br />
23. P.C. Hohenberg: Phys. Rev. 158, 383 (1967)<br />
24. P.M. Chaik<strong>in</strong> and T.C. Lubensky: Pr<strong>in</strong>ciples of condensed matter physics,<br />
(Cambridge University Press, Cambridge 1995)
7 Field-Theory Approaches to Nonequilibrium Dynamics 347<br />
25. S.-k. Ma and G.F. Mazenko: Phys. Rev. Lett. 33, 1383 (1974); Phys. Rev. B11,<br />
4077 (1975)<br />
26. E. Frey and F. Schwabl: Adv. Phys. 43, 577 (1994)<br />
27. H.K. Janssen, B. Schaub, and B. Schmittmann: Z. Phys. B73, 539 (1989)<br />
28. H.K. Janssen: On the renormalized field theory of nonl<strong>in</strong>ear critical relaxation.<br />
In: From phase transitions to chaos, edbyG.Györgyi, I. Kondor, L. Sasvári,<br />
and T. Tél (World Scientific, S<strong>in</strong>gapore 1992), pp. 68–91.<br />
29. P. Calabrese and A. Gambassi: Phys. Rev. E66, 066101 (2002); J. Phys. A:<br />
Math. Gen. 38, R133 (2005)<br />
30. A. Gambassi: In: Proceed<strong>in</strong>gs of the International Summer School “Age<strong>in</strong>g and<br />
the Glass Transition”, to appear <strong>in</strong> J. Phys. Conf. Proc. (July 2006)<br />
31. H.W. Diehl: In: Phase Transitions and Critical Phenomena, vol. 10, ed by<br />
C. Domb and J.L. Lebowitz (Academic Press, London 1986)<br />
32. K. Oerd<strong>in</strong>g and H.K. Janssen: J. Phys. A: Math. Gen. 26, 5295 (1993)<br />
33. U.C. Täuber, V.K. Akk<strong>in</strong>eni, and J.E. Santos: Phys. Rev. Lett. 88, 045702<br />
(2002)<br />
34. F. Haake, M. Lewenste<strong>in</strong>, and M. Wilkens: Z. Phys. B55, 211 (1984)<br />
35. G. Gr<strong>in</strong>ste<strong>in</strong>, C. Jayaprakash, and Y. He: Phys. Rev. Lett. 55, 2527 (1985)<br />
36. K.E. Bassler and B. Schmittmann: Phys. Rev. Lett. 73, 3343 (1994)<br />
37. U.C. Täuber and Z. Rácz: Phys. Rev. E55, 4120 (1997)<br />
38. U.C. Täuber, J.E. Santos, and Z. Rácz: Eur. Phys. J. B7, 309 (1999)<br />
39. B. Schmittmann and R.K.P. Zia: Phys. Rev. Lett. 66, 357 (1991)<br />
40. B. Schmittmann: Europhys. Lett. 24, 109 (1993)<br />
41. K.E. Bassler and Z. Rácz: Phys. Rev. Lett. 73, 1320 (1994); Phys. Rev. E52,<br />
R9 (1995)<br />
42. B. Schmittmann and R.K.P. Zia: Statistical mechanics of driven diffusive systems.<br />
In: Phase Transitions and Critical Phenomena, vol. 17, ed by C. Domb<br />
and J.L. Lebowitz (Academic Press, London 1995)<br />
43. H.K. Janssen and B. Schmittmann: Z. Phys. B63, 517 (1986)<br />
44. K.-t. Leung and J.L. Cardy: J. Stat. Phys. 44, 567 (1986)<br />
45. D. Forster, D.R. Nelson, and M.J. Stephen: Phys. Rev. A16, 732 (1977)<br />
46. J.D. Murray: Mathematical Biology, vols. I/II (<strong>Spr<strong>in</strong>ger</strong>, New York, 3rd ed 2002)<br />
47. M. Mobilia, I.T. Georgiev, and U.C. Täuber: e-pr<strong>in</strong>t q-bio.PE/0508043 (2005)<br />
48. M. Doi: J. Phys. A: Math. Gen. 9, 1465 & 1479 (1976)<br />
49. P. Grassberger and M. Scheunert: Fortschr. Physik 28, 547 (1980)<br />
50. L. Peliti: J. Physique 46, 1469 (1985)<br />
51. F.C. Alcaraz, M. Droz, M. Henkel, and V. Rittenberg: Ann. of Phys. 230, 250<br />
(1994)<br />
52. M. Henkel, E. Orland<strong>in</strong>i, and J. Santos: Ann. of Phys. 259, 163 (1997)<br />
53. G.M. Schütz: In: Phase Transitions and Critical Phenomena, vol. 19, ed by<br />
C. Domb and J.L. Lebowitz (Academic Press, London 2001)<br />
54. R. St<strong>in</strong>chcombe: Adv. Phys. 50, 431 (2001)<br />
55. F. van Wijland: Phys. Rev. E63, 022101 (2001)<br />
56. J.W. Negele and H. Orland: Quantum many-particle systems (Addison-Wesley,<br />
Redwood City 1988)<br />
57. L. Peliti: J. Phys. A: Math. Gen. 19, L365 (1986)<br />
58. H.K. Janssen: J. Stat. Phys. 103, 801 (2001)<br />
59. B.P. Lee: J. Phys. A: Math. Gen. 27, 2633 (1994)<br />
60. B.P. Lee and J. Cardy: Phys. Rev. E50, R3287 (1994)
348 U. C. Täuber<br />
61. D. Toussa<strong>in</strong>t and F. Wilczek: J. Chem. Phys. 78, 2642 (1983)<br />
62. B.P. Lee and J. Cardy: J. Stat. Phys. 80, 971 (1995)<br />
63. O. Deloubrière, H.J. Hilhorst, and U.C. Täuber:Phys.Rev.Lett.89, 250601<br />
(2002); H.J. Hilhorst, O. Deloubrière, M.J. Washenberger, and U.C. Täuber: J.<br />
Phys. A: Math. Gen. 37, 7063 (2004)<br />
64. H.J. Hilhorst, M.J. Washenberger, and U.C. Täuber: J. Stat. Mech. P10002<br />
(2004)<br />
65. M. Moshe: Phys. Rep. 37C, 255 (1978)<br />
66. P. Grassberger and K. Sundermeyer: Phys. Lett. B77, 220 (1978)<br />
67. P. Grassberger and A. De La Torre: Ann. of Phys. 122, 373 (1979)<br />
68. H.K. Janssen and U.C. Täuber: Ann. of Phys. 315, 147 (2005)<br />
69. S.P. Obukhov: Physica A101, 145 (1980)<br />
70. J.L. Cardy and R.L. Sugar: J. Phys. A: Math. Gen. 13, L423 (1980)<br />
71. H.K. Janssen: Z. Phys. B42, 151 (1981)<br />
72. W. K<strong>in</strong>zel: In: Percolation structures and processes,edbyG.Deutsch,R.Zallen,<br />
and J. Adler (Hilger, Bristol 1983)<br />
73. P. Grassberger, Z. Phys. B47, 365 (1982)<br />
74. H. H<strong>in</strong>richsen: Adv. Phys. 49, 815 (2001)<br />
75. G. Ódor:Rev.Mod.Phys.76, 663 (2004)<br />
76. P. Grassberger: Math. Biosc. 63, 157 (1983)<br />
77. H.K. Janssen: Z. Phys. B58, 311 (1985)<br />
78. J.L. Cardy and P. Grassberger: J. Phys. A: Math. Gen. 18, L267 (1985)<br />
79. J. Benzoni and J.L. Cardy: J. Phys. A: Math. Gen. 17, 179 (1984)<br />
80. E. Frey, U.C. Täuber, and F. Schwabl: Europhys. Lett. 26, 413 (1994); Phys.<br />
Rev. E49, 5058 (1994)<br />
81. H.K. Janssen and O. Stenull: Phys. Rev. E62, 3173 (2000)<br />
82. U.C. Täuber, M.J. Howard, and H. H<strong>in</strong>richsen: Phys. Rev. Lett. 80, 2165<br />
(1998); Y.Y. Goldschmidt, H. H<strong>in</strong>richsen, M.J. Howard, and U.C. Täuber: Phys.<br />
Rev. E59, 6381 (1999)<br />
83. R. Kree, B. Schaub, and B. Schmittmann: Phys. Rev. A39, 2214 (1989)<br />
84. F. van Wijland, K. Oerd<strong>in</strong>g, and H. Hilhorst: Physica A251, 179 (1998);<br />
K. Oerd<strong>in</strong>g, F. van Wijland, J.P. Leroy, and H. Hilhorst: J. Stat. Phys. 99,<br />
1365 (2000)<br />
85. J. Cardy and U.C. Täuber: Phys. Rev. Lett. 77, 4780 (1996); J. Stat. Phys.<br />
90, 1 (1998)<br />
86. L. Canet, H. Chaté, and B. Delamotte: Phys. Rev. Lett. 92, 255703 (2004);<br />
L. Canet, H. Chaté, B. Delamotte, I. Dornic, and M.A. Muñoz: e-pr<strong>in</strong>t<br />
cond-mat/0505170 (2005)<br />
87. M. Henkel and H. H<strong>in</strong>richsen: J. Phys. A: Math. Gen. 37, R117 (2004)<br />
88. H.K. Janssen, F. van Wijland, O. Deloubrière, and U.C. Täuber: Phys. Rev.<br />
E70, 056114 (2004)<br />
89. H.K. Janssen: Phys. Rev. E55, 6253 (1997)